MATH 5045: ADVANCED ALGEBRA I (MODULE THEORY)

HEESUNG YANG

1. January 7: Rings
Definition 1.1. A ring R is a set with two binary operations called addition (+) and
multiplication (·) such that
(1) hR, +i is an abelian group
(2) · is associative (i.e., (a · b) · c = a · (b · c) for all a, b, c ∈ R)
(3) · and + are distributive over one another (i.e., a(b+c) = ab+ac and (a+b)c = ac+bc).
Definition 1.2. A ring R is commutative if ab = ba for all a, b ∈ R. Otherwise a ring R is
non-commutative. A ring R has unity (or has identity) if · has an identity, which we call it
1 (i.e., 1 ∈ R and 1 · a = a for all a ∈ R). An element a ∈ R is a unit if there exist a left
multiplicative inverse a0 and a right multiplicative inverse a00 such that a0 a = aa00 = 1.
Example. Z, R, and Z[x] are examples of (commutative) rings. M2 (Z), the 2 × 2-matrix ring
over Z is a (non-commutative) ring.
Proposition 1.1. a0 = a00 . In other words, a left multiplicative inverse of a and a right
multiplicative inverse of a are the same.
Proof. a0 a = 1, so a0 aa00 = a00 . Thus a0 = a00 . 
Definition 1.3. A non-zero element a ∈ R is a zero-divisor if there exists b 6= 0 ∈ R such
that ab = 0 or ba = 0. If R is commutative, has unity, and has no zero-divisors, then R is an
integral domain (or domain in short). A field is an integral domain in which every non-zero
element is a unit.
Example. Z is a commutative ring with unity 1 and units ±1. Z has no zero divisors. Thus Z
is an integral domain. On the other hand, Z/6Z has unity 1 and the units are 1, 5. However,
Z/6Z has three zero divisors, namely 2, 3, 4. Notice that 2 · 3 = 4 · 3 = 0. Therefore Z/6Z is
not an integral domain.
Example. Z/pZ for p prime, Q, R, C, C(x) are examples of fields.
Remark 1.1. Units cannot be zero divisors (left as an exercise).
Definition 1.4. Let R be a ring. A left (resp. right) ideal I of R is a non-empty subset
I ⊆ R such that:
• ra ∈ I (resp. ar ∈ I) for any a ∈ I and r ∈ R
• a − b ∈ I for any a, b ∈ I.
An ideal usually means a left and right ideal.

Date: 9 April 2019.

1
Example. Let R = Z and I = 3Z = {3x : x ∈ Z}. Then I = (3) (i.e., I is an ideal generated
by 3). Since every ideal of Z is generated by a single element, R is in fact a PID (principal
ideal domain). Every ideal is finitely generated in Noetherian rings, so Z is Noetherian.
R[x] is a ring (in fact it is a Euclidean domain). Then (x) and (x2 + 3) are both ideals of
R[x].
Example. However, Z[x] is not a PID (however, it is a UFD (unique factorization domain)).
Note that there does not exist f ∈ Z[x] such that (2, x) = (f (x)).
Definition 1.5. Let R be a ring. A left R-module M over R is an abelian group hM, +i
along with an action of R on M , denoted by multiplication such that
(1) r(x + y) = rx + ry for all r ∈ R and x, y ∈ M
(2) (r + s)x = rx + sx for all r, s ∈ R and x ∈ M
(3) (rs)x = r(sx) for all r, s ∈ R and x ∈ M .
(4) 1R · x = x for all x ∈ M , provided that R has unity.
A right R-module is defined similarly, but with the action of R from the right.
Remark 1.2. Every ring R is an R-module (and a Z-module also).
Example. Every abelian group is a Z-module. Every k-vector space is a k-module for a field
k. Z[x] and Z/6Z are Z-modules.
Example. For every ring R and an ideal I, R/I is an R-module (left as an exercise). Let
r ∈ R and a + I ∈ R/I. Then the action is given by r(a + I) = ra + I.
Example. Let I be an ideal of ring R. Then I is an R-module.

2. January 9
Definition 2.1. Let R be a ring, and M an R-module. Then a submodule of M is a subgroup
N of M which is also an R-module under the same action of R.
Lemma 2.1 (The submodule criterion). Let R be a ring with unity, M a (left) R-module,
and N ⊆ M . Then N is a submodule of N of M if and only if
(1) N is non-empty, and
(2) x + ry ∈ N for any r ∈ R and x, y ∈ N .
Remark 2.1. Notice that R having the unity is crucial, as we will see in the proof. If R has
no unity, then we need to go back to the definition and check one by one instead.
Proof. (⇒) This is a routine application of the definition of an R-module to verify that those
two conditions hold.
(⇐) Suppose that N satisfies the listed criteria. Then N is a subgroup of M . The
first condition implies that there exists x ∈ N . Thus x + (−1)x = 0 ∈ N by the second
condition. Finally, by the second condition, for any x, y ∈ N we have 0 − x = −x ∈ N and
x + 1 · y = x + y ∈ N . Thus for any x ∈ N and r ∈ R, we have 0 + rx = rx ∈ N . Hence N
is closed under action of R. The remaining properties (distributivity) follow because M is
an R-module already: notice that they are inherited from M . 
Definition 2.2. Let R be a ring and M, N R-modules. A function ϕ : M → N is an
R-module homomorphism if
2
 (1) ϕ(x + y) = ϕ(x) + ϕ(y) for all x, y ∈ M
(2L) (for left R-modules) ϕ(rx) = rϕ(x) for all x ∈ M and r ∈ R.
(2R) (for right R-modules) ϕ(xr) = ϕ(x)r for all x ∈ M and r ∈ R.
Additionally, if ϕ : M → N is also
(1) injective, then ϕ is an R-module monomorphism.
(2) surjective, then ϕ is an R-module epimorphism.
(3) bijective, then ϕ is an R-module isomorphism.
(4) M = N , then ϕ : M → M is an R-module endomorphism.
(5) a bijective endomorphism, then ϕ is an R-module automorphism.
Proposition 2.1. ϕ(0) = 0 for any R-module homomorphism ϕ.
Proof. ϕ(0) = ϕ(0 + 0) = 2ϕ(0), so ϕ(0) = 0. 
Example. We examine some examples of module homomorphisms.
• A group homomorphism of abelian groups is a Z-module homomorphism.
• A linear transformation of k-vector spaces is a k-module homomorphism.
• If ϕ : R → S is a ring homomorphism, then S is an R-module with action of R
defined as r · x = ϕ(r)x for all r ∈ R, x ∈ S. Then S is an R-module. Evidently, R
is also an R-module, so ϕ is in fact an R-module homomorphism. Indeed,
(1) ϕ(x + y) = ϕ(x) + ϕ(y) for all x, y ∈ R (since ϕ is a ring homomorphism)
(2) ϕ(rx) = ϕ(r)ϕ(x) = r · ϕ(x) = rϕ(x) for r, x ∈ R.
Lemma 2.2. Let R be a ring with unity, and M and N are left R-modules. Then the
following are equivalent:
(i) ϕ : M → N is an R-module homomorphism.
(ii) ϕ(x + ry) = ϕ(x) + rϕ(y) for all x, y ∈ M and r ∈ R.
Proof. Exercise. 
Definition 2.3. Let ϕ : M → N be a homomorphism of left R-modules. Then kernel of ϕ
is
ker ϕ = {x ∈ M : ϕ(x) = 0}.
The image of ϕ is
im ϕ = {y ∈ N : y = ϕ(x) for some x ∈ M }.
Lemma 2.3. If ϕ : M → N is a left R-module homomorphism, then ϕ(M ) = im ϕ is
submodule of N , and ker ϕ is submodule of M .
Proof. From group theory, we already know that ker ϕ and im ϕ are subgroups. Thus we
only need to verify they are also modules. For ϕ(M ), for any r ∈ R and x ∈ ϕ(M ) there
exists y ∈ M such that x = ϕ(y). Thus, rx = rϕ(y) = ϕ(ry) ∈ ϕ(M ) since ry ∈ M . Thus
ϕ(M ) is a submodule of N .
As for the kernel, for any r ∈ R and x ∈ ker ϕ we have ϕ(rx) = rϕ(x) = r0 = 0. Thus
rx ∈ ker ϕ, as required. 
Definition 2.4. Let M, N be left R-modules, and let
HomR (M, N ) := {ϕ : M → N | ϕ is an R-module homomorphism}.
3
Define addition on HomR (M, N ) as follows. For any ϕ, ψ ∈ HomR (M, N ), define
(ϕ + ψ)(x) := ϕ(x) + ψ(x) for all x ∈ M.
It is not hard to see that ϕ + ψ : M → N is an R-module homomorphism. We see ϕ + ψ
respects addition since for any x, y ∈ M ,
(ϕ + ψ)(x + y) = ϕ(x + y) + ψ(x + y)
= ϕ(x) + ϕ(y) + ψ(x) + ψ(y)
= (ϕ + ψ)(x) + (ϕ + ψ)(y).
Similarly, we have, for any r ∈ R and x ∈ M ,
(ϕ + ψ)(rx) = ϕ(rx) + ψ(rx) = rϕ(x) + rψ(x)
= r(ϕ(x) + ψ(x)) = r((ϕ + ψ)(x)).
Hence ψ + ϕ ∈ HomR (M, N ) for all ϕ, ψ ∈ HomR (M, N ). Let 0 ∈ HomR (M, N ) be the zero
homomorphism 0 : M → N (i.e., 0(x) = 0 for all x ∈ M ), which serves as the identity
element. It is not that hard to see that −ϕ ∈ HomR (M, N ) defined as x 7→ −ϕ(x) is also an
R-module homomorphism for any ϕ ∈ HomR (M, N ). Therefore ϕ + (−ϕ) = 0.
Thus, we show that hHomR (M, N ), +i is an abelian group. Can we make HomR (M, N )
into an R-module? The answer is yes, provided that R is commutative, with action of R
defined as (rϕ)(x) = rϕ(x) = ϕ(rx) for any r ∈ R, x ∈ M, ϕ ∈ HomR (M, N ).

3. January 11
Let R be a commutative ring, M, N R-modules. We define an action of R on HomR (M, N )
as follows: let rϕ : M → N satisfy (rϕ)(x) = rϕ(x) where ϕ is an R-module homomorphism
from M to N . We need to verify that rϕ : M → N is an R-module homomorphism.
(1) (rϕ)(x + y) = r · ϕ(x + y) = r(ϕ(x) + ϕ(y)) = r · ϕ(x) + r · ϕ(y) = (rϕ)(x) + (rϕ)(y)
fo rall x, y ∈ M and r ∈ R.
(2) Let r, s ∈ R and x ∈ M . Then (rϕ)(sx) = r · ϕ(sx) = rsϕ(x) = srϕ(x) = s(rϕ)(x),
as needed.
Proposition 3.1. HomR (M, N ) under the action of R defined above is an R-module.
Proof. We know HomR (M, N ) is an abelian group and is closed under the action. So it
remains to verify the criteria for modules. Suppose that r, s ∈ R and ϕ, ψ ∈ HomR (M, N ).
(1) We need to show that (r + s)ϕ = rϕ + sϕ. (Exercise)
(2) We need to show that r(ϕ + ψ) = rϕ + rψ. (Exercise)
(3) We also need to show that (rs)ϕ = r(sϕ). Indeed, ((rs)ϕ)(x) = rsϕ(x) = r(sϕ(xx)) =
r(sϕ)(x).
Thus HomR (M, N ) is an R-module as required. 

3.1. Composition of homomorphisms

Proposition 3.2. Let M, N, L be R-modules, and suppose ϕ ∈ HomR (M, L) and ψ ∈
HomR (L, N ). Then ψ ◦ ϕ : M → N ∈ HomR (M, N ), i.e., ψ ◦ ϕ is a homomorphism.
4
Proof. This is a straightforward verification.
ψ ◦ ϕ(x + y) = ψ(ϕ(x + y)) = ψ(ϕ(x) + ϕ(y)) = ψ ◦ ϕ(x) + ψ ◦ ϕ(y)
ψ ◦ ϕ(rx) = r(ψ ◦ ϕ(x))(Exercise.),
since ψ and ϕ are R-module homomorphisms. 
Proposition 3.3. Suppose R is a commutative ring and M an R-module. Let + be the
usual addition, and · be the composition of homomorphisms. Then HomR (M, M ) is a ring
with unity 1.
Proof. Exercise. 

3.2. Quotient modules

Suppose M is an R-module, and N a submodule of M . Then M/N is the quotient group
{x + N : x ∈ M }. Notice that R can act on M/N . For any r ∈ R and x + N ∈ M/N , let
the action be
r(x + N ) := rx + N.
First, observe that this action is well-defined. Indeed, if x + N = y + N in M/N , and r ∈ R,
then x − y ∈ N . But N is a submodule, so r(x − y) ∈ N also. Hence rx − ry ∈ N so
rx + N = ry + N , as required. Second, we want to show that M/N is an R-module under
this action. That is, we need to verify the three following conditions:
(1) r((x + y) + N ) = (rx + N ) + (ry + N ) (Exercise)
(2) (r + s)(x + N ) = r(x + N ) + s(x + N )
(3) (rs)(x + N ) = r(sx + N )
Definition 3.1. The (group) projection map π : M → M/N is defined by π(x) = x + N .
It is evident that π is a(n additive) group homomorphism. That π is R-linear is also evident:
for any r ∈ R and x ∈ M , we have π(rx) = rx + N = r(x + N ) = rπ(x).

3.3. Isomorphism theorems for modules

Assume that M, N are R-modules, and that A and B are submodules of M .
Theorem 3.1 (First isomorphism theorem for modules). Let ϕ : M → N be a R-module
homomorphism. Then ker ϕ is a submodule of M and M/ ker ϕ ∼
= ϕ(M ).
Proof. First part: Exercise. Since M/ ker ϕ ∼ = ϕ(M ) as groups already, by the first isomor-
phism theorem for groups, it suffices to verify that the group isomorphism given by the first
isomorphism theorem for groups is R-linear. (Exercise.) 
Theorem 3.2 (Second isomorphism theorem for modules). (A + B)/B ∼
= A/(A ∩ B).
Proof. Pick an appropriate ϕ : A + B → A/(A ∩ B). Show that ϕ is surjective and that
ker ϕ = B. Just show that ϕ is R-linear, and then apply the first isomorphism theorem. Do
not try to show that the map is additive – this is already given by the theorem for group
counterparts. 
Theorem 3.3 (Third isomorphism theorem for modules). If A ⊆ B, then (M/A)/(M/B) ∼
=
A/B.
5
Theorem 3.4 (Correspondence theorem for modules (Fourth isomorphism theorem for mod-
ules)). There is an inclusion-preserving one-to-one correspondence between the set of sub-
modules of M containing A and the set of submodules of M/A. This correspondence com-
mutes with taking sums and intersections (i.e., there is an isomorphism of lattices between
the submodule lattice of M/A and the lattice of submodules of M containing A).
Remark 3.1. The last statement of the fourth isomorphism theorem for modules shows why
the theorem is also called the “lattice isomorphism theorem”.

4. January 14
Definition 4.1. A category is a collection of objects and morphisms between the objects.
A category C comes with:
• Obj(C): collection of objects in C.
• for every A, B ∈ Obj(C) a set HomC (A, B) of morphisms f : A → B with domain A
and codomain B of f such that:
(i) for every A ∈ Obj(C) there exists 1A ∈ HomC (A, A) which is the identity mor-
phism on A. Therefore, there is always a morphism in HomC (A, A) = EndC (A) 6=
∅ (endomorphisms).
(ii) f ∈ HomC (A, B) and g ∈ HomC (B, C) give a morphism gf ∈ HomC (A, C).
Hence, there exists a set function
HomC (A, B) × HomC (B, C) → HomC (A, C)
(f, g) 7→ gf.
(iii) Composition is associative: f ∈ HomC (A, B), g ∈ HomC (B, C), h ∈ HomC (C, D),
then h(gf ) = (hg)f .
(iv) For every f ∈ HomC (A, B), f 1A = f and 1B f = f .
(v) If HomC (A, B) ∩ HomC (C, D) 6= ∅, then A = C and B = D.
4.1. Generators for modules
Let R be a ring with unity 1. Let M be an R-module, and N1 , N2 , . . . , Nk submodules of
M.
Definition 4.2. The sum of N1 , . . . , Nk is
N1 + N2 + · · · + Nk := {x1 + · · · + xk | xi ∈ Ni for all i}.
Proposition 4.1. N1 + · · · + Nk is a submodule of M .
Proof. Exercise. 
Remark 4.1. If N1 , . . . , Nk are submodule of N , then N1 + · · · + Nk is a submodule of M
generated by N1 ∪ · · · ∪ Nk .
Definition 4.3. Let A ⊆ M be a subset (not necessarily a submodule). Then define
RA := {r1 a1 + · · · + rn an : a1 , . . . , an ∈ A, r1 , . . . , rn ∈ R},
which generates a submodule. We call RA the submodule of M generated by A (the smallest
submodule of M containing A). If A = ∅ we say RA = {0}. If A is finite, then RA is finitely
generated. If |A| = 1, then RA is a cyclic module.
6
 It is not entirely obvious if RA is actually a module, but it is not a difficult exercise to
prove this is indeed the case.
Proposition 4.2. RA is indeed a submodule of M .
Proof. Exercise. 
Example. R is a cyclic R-module because R = R1R . R/I is another example of a cyclic
R-module since R/I = R(1R + I). Z[x]/(x2 ) = h1, xi as a Z-module. However, Z[x] is not a
finitely generated Z-module, since Z[x] is generated by {1, x, x2 , x3 , . . . }.
Definition 4.4. If M1 , . . . , Mk are R-modules, then the direct product of M1 , . . . , Mk is the
collection
Yk
Mi = M1 × M2 × · · · × Mk = {(m1 , . . . , mk ) : mi ∈ Mi ∀i}.
i=1

This is also called the external direct sum of M1 , . . . , Mk , denoted by M1 ⊕ M2 ⊕ · · · ⊕ Mk .

Remark 4.2. For a family of abelian groups {Gi : i ∈ I} (note that I may be uncountable),
the direct product and the direct sum as follows:
Y n [ o
Gi = f : I → Gi | f (i) ∈ Gi ∀i ∈ I
i∈I
X n Y o
Gi = f ∈ Gi | f (i) = 0 for all but finitely many i ∈ I .
i∈I
Q S
For any f, g ∈ Gi , define the composition f g : I → Gi be i 7→ f (i) + g(i). Therefore, if I
is finite, then the direct sum
Q and the direct product are equal. Finally, it is a straightforward
verification to check that Gi is a group.
Proposition 4.3. M1 × · · · × Mk is an abelian group under component-wise addition. Fur-
thermore, we can define a component-wise action on R
r(x1 , . . . , xk ) = (rx1 , . . . , rxk ),
making M1 × · · · × Mk into an R-module.
Proposition 4.4 (Direct sum of submodules). Let R be a ring with unity and M an R-
module. Let N1 , . . . , Nk be submodules of M . Then the following are equivalent:
(i) The map π : N1 × · · · × Nk → N1 + · · · + Nk defined by
(n1 , . . . , nk ) 7→ n1 + · · · + nk
is an isomorphism of R-modules.
(ii) Nj ∩ (N1 + · · · + Nj−1 + Nj+1 + · · · + Nk ) = {0} for all j ∈ {1, 2, . . . , k} (mod k).
(iii) For any x ∈ N1 + · · · + Nk , x can be written uniquely as a1 + · · · + ak where ai ∈ Ni .
Definition 4.5. If N1 + · · · + Nk satisfies any of the conditions listen in Proposition 4.4,
then N1 + · · · + Nk is the internal direct sum of N1 , . . . , Nk , and we write N1 ⊕ N2 ⊕ · · · ⊕ Nk .
7
 5. January 16
X
Proof of Proposition 4.4. ((1) ⇒ (2)) If Nj ∩ Ni contains an element aj 6= 0, then there
i6=j
exists ai ∈ Ni where i 6= j such that
X
aj = ai .
i6=j

So a1 + · · · + aj−1 − aj + aj+1 + · · · + ak = 0. So if π((a1 , . . . , ak )) = 0, then a1 = · · · = ak = 0.

Thus aj = 0, but it is a contradiction.
((2) ⇒ (3)) Suppose that a1 + · · · + ak = b1 + · · · + bk . Then there exist ai , bi ∈ Ni where
i = 1, 2, . . . , k. Fix j ∈ {1, 2, . . . , k}, and one can write
X
aj −bj = (b1 −a1 )+(b2 −a2 )+· · ·+(bj−1 −aj−1 )+(bj+1 −aj+1 )+· · ·+(bk −ak ) ∈ Nj ∩ Ni = 0.
i6=j

Thus aj − bj = 0, so aj = bj for every j as required.

((3) ⇒ (1)) Let π : N1 ×· · ·×Nk → N1 +· · ·+Nk is an isomorphism because π(a1 , . . . , ak ) =
0 implies a1 + · · · + ak = 0. Thus a1 = a2 = · · · = ak = 0. Therefore π is injective. Clearly,
π is surjective (clear from the definition of π). Also, it is straightforward to verify that π is
a module homomorphism, so this will be left as an exercise. 
5.1. Universal property of direct sum of modules
Theorem 5.1. Let R be a ring, let {Mi | i ∈ I} be a family of R-modules, N an R-module,
and {ψi : Mi → N | i ∈ I} a family of R-module homomorphisms. Then there exists a
unique R-module homomorphism X
ψ: Mi → N
i∈I
P
such that ψi = ψMi for all i ∈ I. P Furthermore, this Mi is uniquely determined up to
isomorphism by this property (i.e., Mi is a co-product in the category of R-modules).
Proof. It is known that this works for all groups – we can define
X
ψ: Mi → N
i∈I
P
by ψ((ai )i∈I ) = ψi (ai ). Verify that this is a group homomorphism and is R-linear (exer-
cise). Also, it is a routine exercise to verify the rest of the claims. 
5.2. Exact sequences
Definition 5.1. Let M, N, L be R-modules. Then the sequence of R-module homomor-
phisms
f g
M →N →L
is called exact at N if f is injective, g is surjective, and im g = ker f . Similarly, a long exact
sequence is
fi fi+1
· · · → Mi−1 → Mi → Mi+1 → · · ·
such that for every Mi , ker fi+1 = im fi for all i. A short exact sequence is of the form
f g
0→M →N →P →0
8
such that f is injective, g is surjective, and im f = ker g.
f g
Remark 5.1. If 0 → M → N is exact at M , then ker g = im f = 0. Therefore g is injective.
f g
Similarly, if M → E → 0 is exact at N , so ker g = N = im f . Thus f is surjective in this
case.
i
Example. If M is an R-module and N a submodule of M , then 0 → N → M is exact;
π
similarly, M → N → 0 is exact as well. Thus we get the short exact sequence
i π
0 7→ N → M → M/N → 0
where i is the injection map and M the projection map.
Definition 5.2. The co-kernel of an R-module homomorphism f : M → N is CoKer(f ) :=
N/ im f .
Remark 5.2. Let f : M → N be an R-module homomorphism. Then we have an exact
sequence
f π
0 → ker f → M → N → CoKer(f ) → 0.
How many short exact sequences can we extract out of this? We can generate at least two
short exact sequences. 0 → ker f → M → im f → 0 and 0 → im f → N → N/ im f → 0.
Example. For any M , N , and their direct sum M ⊕ N , the sequence
i π
0→M →M ⊕N →N →0
is a short exact sequence. Note that im i = M ⊕ 0, and clearly ker ϕ = M ⊕ 0.
6. January 18
Definition 6.1. Suppose that
f g
0→A→B→C→0
is a short exact sequence. Then this short exact sequence is split exact if B ∼
= A ⊕ C.
Definition 6.2. Two short exact sequences 0 → A → B → C → 0 and 0 → A0 →
B 0 → C 0 → 0 of R-modules are isomorphic if there is a commutative diagram of R-module
homomorphisms such that g ◦ α = α0 ◦ f and h ◦ β = β 0 ◦ g.
β
0 / A
α / B / C / 0
f g h
  β0

/ α0 / / /
0 A0 B0 C0 0
Theorem 6.1. Let R be a ring, and let 0 → A → B → C → 0 be a short exact sequence of
R-module. Then the following are equivalent:
(i) There exists an R-module homomorphism h : C → B such that g ◦ h = idC .
(ii) There exists an R-module homomorphism k : B → A such that k ◦ f = idA .
(iii) B ∼
= A ⊕ C and the sequence above can be isomorphically written as
1 i π
0→A→ A ⊕ C →2 C → 0.
Therefore the short exact sequence is split exact.
9
 To prove the equivalent conditions for split exact sequence, we need the following lemma.

Lemma 6.1 (Short five lemma). Let R be a ring, and where is a commutative diagram of
R-modules and R-module homomorphisms
f g
0 / A / B / C / 0
α β γ
 f0
 g0

0 / A0 / B0 / C0 / 0
such that each row is a short exact sequence. Then
(i) If α and γ are monomorphisms, then β is also a monomorphism.
(ii) If α and γ are epimorphisms, then β is also an epimorphism.
(iii) If α and γ are isomorphisms, then β is also an isomorphism.
Proof. (i) Suppose x ∈ ker β. Then β(x) = 0, so (g 0 ◦ β)(x) = 0. But then g 0 ◦ β = γ ◦ g. But
then γ is a monomorphism, so g(x) = 0. Hence x ∈ ker g = im f . So there exists y ∈ A such
that x = f (y). Hence (β ◦ f )(y) = (f 0 ◦ α)(y) = 0; but f 0 is a monomorphism, so α(y) = 0.
But again α is also a monomorphism, so y = 0. Hence x = f (y) = 0 as needed.
(ii) Let y ∈ B 0 . Then g 0 (y) ∈ C 0 . But since γ is an epimorphism, there exists z ∈ C
such that g 0 (y) = γ(z). But g is an epimorphism, so there is u ∈ B such that z = g(u).
So g 0 (y) = γ(z) = (γ ◦ g)(u) = (g 0 ◦ β)(u). It thus follows that g 0 (β(u) − y) = 0, so
β(u) − y ∈ ker g 0 = im f 0 . Since β(u) − y ∈ im f 0 , there is v ∈ A0 such that β(u) − y = f 0 (v).
α is an epimorphism, so one can find w ∈ A such that β(u) − y = (f 0 ◦ α)(w) = (β ◦ f )(w).
So β(u − f (w)) = y. This proves that β is surjective.
(iii) This is immediate from (i) and (ii). 
Proof of Theorem 6.1. ((i) ⇒ (iii)) Consider the two short exact sequences
f g
0 A B C 0
h
id ϕ id
ι1 π2
0 A A⊕C C 0

We need to show that these two sequences are isomorphic. Thus we need to find an
isomorphism ϕ such that the diagram above commutes. Define ϕ : A ⊕ C → B by
(a, c) 7→ f (a) + h(c). Note that ϕ is well-defined since (a, c) is a unique representative
for this element, and both f and h are well-defined. ϕ is a homomorphism since
ϕ(r(a, c)) = ϕ((ra, rc)) = f (ra) + h(rc) = r(f (a) + h(c)) = rϕ(a, c)
ϕ((a, c) + (a0 , c0 )) = ϕ((a + a0 , c + c0 )) = f (a + a0 ) + h(c + c0 )
= f (a) + h(c) + f (a0 ) + h(c0 ) = ϕ((a, c)) + ϕ((a0 , c0 )).
We want to show that the diagram commutes. Pick (a, c) ∈ A ⊕ C. Then (g ◦ ϕ(a, c) =
g(f (a) + h(c)) = (g ◦ f )(a) + (g ◦ h)(c) = c. On the other hand, (id ◦π2 )(a, c) = id(c) = c.
Thus g ◦ ϕ ≡ id ◦π2 . We can use the similar argument to show that the other side commutes,
i.e., ϕ ◦ i1 ≡ f ◦ id. That ϕ is an isomorphism follows from the short five lemma.
10
 ((ii) ⇒ (iii)) Assume that there is k such that k ◦ f = idA . Define ϕ : B → A ⊕ C so
that b 7→ (k(b), g(b)). ϕ is well-defined since k and g are well-defined also. ϕ is also an
R-module homomorphism since k and g are. Indeed, ϕ(b1 + b2 ) = (k(b1 + b2 ), g(b1 + b2 )) =
(k(b1 ), g(b1 )) + (k(b2 ), g(b2 )) = ϕ(b1 ) + ϕ(b2 ); also for any r ∈ R, ϕ(rb1 ) = (k(rb1 ), g(rb1 )) =
(rk(b1 ), rg(b1 )) = r(k(b1 ), g(b1 )) = rϕ(b1 ). So by the short five lemma, ϕ is an isomorphism,
so the two short exact sequences are isomorphic as desired.
((iii) ⇒ (i), (ii)) We have an isomorphism of short exact sequences, i.e., ϕ1 , ϕ2 , and ϕ3 are
all isomorphisms.
f g
0 A B C 0
k h
id ϕ id
ι1 π2
0 A A⊕C C 0
π1 ι2

We let h : C → B where h = ϕ−1 2 i2 ϕ3 . Note that h is well-defined since it is just the

composition of three homomorphisms. For any c ∈ C, observe that ϕ−1 2 i2 ϕ3 (c) ∈ B. So by
the commutativity, ϕ3 g(b) = π2 ϕ2 (b) = π2 ϕ2 (ϕ−1 i
2 2 3ϕ (c)) = π (i ϕ
2 2 3 (c)) = ϕ3 (c). But then
ϕ3 is an isomorphism, so g(b) = c from which gh(c) = c follows. Hence gh = idC .
Now define k : B → A by k := ϕ−1 1 π1 ϕ2 which is a well-defined homomorphism for the
same reason h is. For any a ∈ A, we have kf (a) = ϕ−1 −1
1 π1 ϕ2 f (a) = ϕ1 π1 i1 ϕ1 (a) = a, as
desired. 
Remark 6.1. If M a R-module and M1 , M2 submodules of M , we have a short exact sequence
f g
0 −→ M1 ∩ M2 −→ M1 ⊕ M2 −→ M1 + M2 −→ 0,
where f : m 7→ (m, −m) and g : (m1 , m2 ) 7→ m1 + m2 .

7. Detour: Nakayama’s lemma

Definition 7.1. Let R be a commutative ring with unity. If R has a unique maximal ideal
m, then (R, m) is a local ring.
Lemma 7.1. Let R be a ring, I an ideal of R, and M an R-module. Then
IM = {am | a ∈ I, m ∈ M }
is a submodule of M .
Proof. Exercise. 
Lemma 7.2. If M is a R-module, and I an ideal of R, then M/IM is an R/I-module,
where the action of R/I is defined by (r + I)(x + IM ) : f = rx + IM .
Proof. Exercise. 
Remark 7.1. Recall that if (R, m) is a local ring, then the only non-units of R are precisely
the elements of m. Suppose that is not the case. Pick x ∈ R \ m. Consider the ideal I = (x),
and that 1 ∈/ I (since x is not a unit). Thus I 6= R. Since m is the only maximal ideal, it
follows that (x) ≤ m. But this means x ∈ m which is a contradiction.
11
Theorem 7.1 (Nakayama’s lemma). Let R be a commutative ring with unity 1, I be an ideal
of R, and M a finitely generated R-module. If IM = M , then there exists r ∈ R satisfying
r ≡ 1 (mod I) that vanishes M (i.e., rM = 0).
Theorem 7.2 (Nakayama’s lemma, local ring version). Let (R, m) be a local ring, and M
an R-module. Suppose that x1 , . . . , xn ∈ M . Then the following are equivalent:
(i) M = hx1 , x2 , . . . , xn i is a finitely generated R-module.
(ii) M/mM = hx1 , x2 , . . . , xn i is an R/m-vector space (xi is the image of xi under the
map M → M/mM . Note that R/m is a field, so any R/m-module is automatically
an R/m-vector space.
Proof. (⇒) this direction is straightforward from the definition.
(⇐) Let N = hx1 , . . . , xn i. We want to show that M/N = 0. We can rephrase this
problem: we can instead show that if M is finitely generated and M/mM = 0 then M = 0.
We will prove this claim by induction on the number of generators.
Since M is finitely generated, there exist y1 , y2 , . . . , yt ∈ M such that M = hy1 , . . . , yt i. If
t = 1 then M = hy1 i and M = mM = my1 . Thus there is a ∈ m such that y1 = ay1 . Then
(1 − a)y1 = 0. Note that 1 − a ∈ / m (since a ∈ m 6= R), so 1 − a is a unit. Hence y1 = 0,
whence we have M = hy1 i = 0.
Suppose t > 1, and that M = mM . Then there exist a1 , . . . , at ∈ m so that yt =
a1 y1 + · · · + at yt . Then (1 − at )yt = a1 y1 + · · · + at−1 yt−1 . Then 1 − at ∈ / m, so 1 − at
−1 −1
is a unit. Hence yt = a1 (1 − at ) y1 + · · · + at−1 (1 − at ) yt−1 ∈ hy1 , . . . , yt−1 i. Thus
M = hy1 , . . . , yt i = hy1 , . . . , yt−1 i. Thus we can induct on t to reduce it to the base case.
The claim follows. 

8. January 23: Free modules

Suppose that M is an R-module where R is a ring with unity 1.
Definition 8.1. A subset R of M is called linearly independent if a1 x1 + · · · + an xn = 0
implies a1 = a2 = · · · = an = 0 for all a1 , . . . , an ∈ R and x1 , x2 , . . . , xn ∈ X. If M is
generated by a linearly independent subset X, then X is called a basis of M . A free module
is a module with a non-empty basis.
Theorem 8.1. Suppose that R is a ring with identity, and F an R-module. Then the
following are equivalent:
(i) F has a non-empty basis.
(ii) F is the internal direct sum of cyclic submodules.
(iii) F is isomorphic to a direct sum of copies of R (i.e., F ∼
= Rn for some n; alternatively,
F ∼
L
= R.)
Proof. ((ii) ⇔ (iii)) They are equivalent statements since Rx ∼ = R for any non-zero x ∈ X.
((i) ⇒ (ii) & (iii)) If X 6= ∅ is a basis of F and x ∈ X, then we have a surjective R-module
homomorphism ϕx : R → Rx defined by ϕx (r) := rx. ϕx is injective, since if rx = 0 then
r = 0 (note that x ∈ X is a basis, so x 6= 0). Thus ker ϕx = 0 as needed. It is not hard to
check that ϕx is a homomorphism.
Hence, we have
F ∼ Rx ∼
M M
= = R.
x∈X x∈X
12
Note that the second direct sum is internal, whereas the third direct sum is external; note
also that the second isomorphism follows since ϕx is an isomorphism (and replace each Rx
with R).
Ψ M
((iii) ⇒ (i)) Suppose that F ∼
= R where X is the index set of this direct sum. Define
x∈X
ιx ∈ F to be the tuple such that
(
1 (x = y)
(ιx )y =
0 otherwise.
M
Then {ιx : x ∈ X} is a basis for R. The image of {ιx : x ∈ X} under Ψ is a basis for F .
x∈X


9. January 25
Definition 9.1. A division ring (or a skew field ) is a ring with 1 such that every non-zero
element in a unit. A field is a commutative division ring, and a vector space is a module
over a division ring.
Example. The quaternion ring is a standard example of a division ring.
Lemma 9.1. Let V be a vector space over a division ring D, and let X be a maximal linearly
independent subset of V . Then X is a basis of V .
Proof. If V 0 = hXi ⊆ V , we want to show that V 0 = hV i. Since X is linearly independent, it
is a basis of V 0 . Let x ∈ V \ V 0 . Then X ∪ {x} is linearly independent. Suppose otherwise.
Then if
d1 x1 + · · · + dn xn + dx = 0
where di , d ∈ D and xi ∈ X, we have
x = d−1 (d1 x1 + · · · + dn xn ) ∈ V 0 .
/ V 0 . This forces d = 0, so d1 x1 + · · · + dn xn = 0. In turn,
But this is a contradiction since x ∈
this implies d1 = d2 = · · · = dn = 0 as well. This implies X ∪ {x} is linearly independent,
but this contradicts the fact that X is a maximal linearly independent set. 
Theorem 9.1 (Zorn’s lemma). Let A 6= ∅ be a partially ordered set, such that every chain
has an upper bound in A. Then A contains a maximal element.
Theorem 9.2. Let V be a vector space over a division ring D. Then V has a basis, so V
is a free D-module. Moreover, if Y is a linearly independent subset of V , then there exists a
basis X of V such that Y ⊆ X.
Proof. The first part follows from the second part, and clearly ∅ is (vacuously) linearly
independent by default, so wle will prove the second part only. Let
A := {X ⊆ V : X linearly independent and Y ⊆ X}.
Since Y ∈ A, A 6= ∅. A is partially ordered by inclusion. If C is a chain in A, define
[
X := X ∈ A.
X∈C
13
Then X is an upper bound of C. By Zorn’s lemma, A contains a maximal element B, so by
Lemma 9.1, B is a basis of V . 
Theorem 9.3. If V is a vector space over a division ring D, then every generating set of V
contains a basis of V .
Proof. If X is a generating set of V , let A := {Y | Y ⊆ X linearly independent}, which is
a partially ordered set under inclusion. Again, every chain has an upper bound by Zorn’s
lemma. Suppose that Y is a maximal element of A. Then x ∈ hY i for all x ∈ X (otherwise,
we can add an element to Y , which contradicts the maximality of Y ). Hence V ⊆ hXi ⊆ hY i,
so V = hY i. 

10. January 28 & 30

Theorem 10.1. Let X be any set, and R a ring with unity. Then there exists a free R-
module F (X) on X satisfying the following universal property: for any R-module M and
ϕ : X → M a function, there is a unique R-module homomorphism Φ : F (X) → M such
that Φ(x) = ϕ(x) for all x ∈ X. In other words, the following diagram commutes.
ϕ
X M
ι ∃!Φ

F (X)

Proof. Build F (X). If X = ∅ then F (X) = 0. Otherwise, F (X) = {f : X → R : f (x) =

0 for all but finitely many x ∈ X}. We will make F (X) into an R-module. Let f, g ∈ F (X)
and r ∈ R, and let
(f + g)(x) := f (x) + g(x)
(rf )(x) := r.f (x)
for all x ∈ X. If x ∈ X define fx ∈ F (X) as
(
1 y=x
fx (y) :=
0 otherwise.
So if f ∈ F (X) then there are x1 , . . . , xn ∈ X such that
f = f (x1 )fx1 + · · · + f (xn )fxn .
Note that f (xi ) ∈ R and fxi ∈ F (X) for all i. And we know this is unique, so {fx : x ∈ X}
is a basis for F (X). Thus F (X) is a free R-module.
To check the universal property, suppose ϕ : X → M . Define Φ : F (X) → M so that
n
! n
X X
Φ ai f x i = ai ϕ(xi ).
i=1 i=1

It is not hard to check if it is well-defined, is a homomorphism, and Φ|X = ϕ (Exercise).

Every element of F (X) has a unique presentation in the form of
X n
ai fxi
i=1
14
for some n ∈ Z+ , ai ∈ R, and xi ∈ X. Thus Φ is the unique extension of ϕ to F (X) as
needed. 
Proposition 10.1. Every finitely generated R-module for R a ring with identity is the
homomorphic image of a finitely generated free module.
Proof. Let X := {x1 , . . . , xn }, and M = hXi be a finitely generated R-module. By the
universal property, there is a free R-module F (X) and a homomorphism ϕ : F (X) → M
satisfying fx 7→ x. 
Remark 10.1. In fact, M ∼ = F (X)/ ker ϕ ∼
= Rn / ker ϕ.
10.1. Free modules and ranks
Suppose that F is a free module over a ring with 1. Do every two bases necessarily have
the same cardinality? The answer is actually no in general, but it is true for commutative
rings and division rings. Our main goal in this section is to prove this is indeed the case.
Definition 10.1. Let R be a commutative ring or a division ring, and let X be a basis of a
free R-module F . Then the rank of F is the cardinality of X.
Theorem 10.2. Let R be a ring with unity, and F a free module with basis X with |X| = ∞.
Then every basis of X has the same cardinality as X. Therefore, if the basis is infinite, then
the cardinality is unique regardless of what the ring is.
Proof. Suppose Y is another basis of F whose basis is X. If Y is finite, suppose Y =
{y1 , . . . , yn }. Then for all yi ∈ Y one can find xi,1 , . . . , xi,mi ∈ X and ri,1 , . . . , ri,mi ∈ R so
that yi = ri,1 xi,1 + · · · + ri,mi xi,mi . Then X 0 = {xi,j : 1 ≤ i ≤ n, 1 ≤ j ≤ mi } is a finite subset
of X spanning F . Therefore X contains a finite-generating set for F , but this contradicts
the fact that |X| = ∞. Therefore |Y | is infinite.
Let K(Y ) be the set of finite subsets of Y , and define f : X → K(Y ) so that x 7→
Xn
{y1 , . . . , yn } where x = ri yi is uniquely defined. (i.e., r1 , r2 , . . . , rn ∈ R \ {0} are unique,
i=1
and y1 , . . . , yn ∈ Y are uniquely determined by x. Therefore f is well-defined. We make a
few observations regarding f . [
First, im f is an infinite set. Suppose otherwise, and let X = h Ai. Note that
A∈im f
A = f (x) for some x. Thus A is a finite set, and the finite union of finite sets is finite. Thus
F is generated by a finite subset of Y , which is a contradiction. Second, for any S ∈ im f
we have |f −1 (S)| < ∞. Let x ∈ f −1 (S). Then x ∈ hy : y ∈ Si is a submodule of F . Hence
f −1 (S) ⊆ hy : y ∈ Si. Each y in S thus can be uniquely written as a sum of finite elements
of X, and |S| < ∞. Hence f −1 (S) ⊆ hXS i, where XS is a finite subset of X.
−1
PNow, if x ∈ f−1 (S), then there are x1 , .−1 . . , xn ∈ XS and r1 , . . . , rn ∈ R such that x =
Ri xi . Thus f (S) ⊂ XS . Therefore |f (S)| ≤ |XS | < ∞. Now let s ∈ im(f ). Then,
say, f −1 (S) = {x1 , . . . , xn }. Define gS : f −1 (S) → im f × N by xi 7→ (S, i). Now we claim
that the sets f −1 (S) for S ∈ im f forms a partition of X. It is a relatively straightforward
exercise to verify that [
X= f −1 (S),
S∈im f
and if x ∈ X, there exists a unique {y1 , . . . , yn } = S ⊆ Y such that x ∈ hy1 , . . . , yn i.
15
 Thus define g : X → im f × N by x 7→ gS (x) where x ∈ f −1 (S). Note that g is well-defined
and injective. Furthermore, |X| ≤ | im f | × |N| = | im f |ℵ0 = | im f | ≤ |K(Y )| = |Y | (For
more information, refer to Hungerford’s I.8.13).
Now use the reverse argument to show that |Y | ≤ |X|, from which |X| = |Y | follows. 
Corollary 10.1. Let V be a vector space over a division ring D, and X, Y two bases of V .
Then |X| = |Y |.
Now that we got the infinite case out of the way, we can move on to the finite basis case.
Recall that we claimed that the rank of a free R-module is well-defined only when R is a
division ring or a commutative ring.
Theorem 10.3. Let V be a finite-dimensional vector space over a division ring D. Let X
and Y be two bases of V . Then |X| = |Y |.
Proof. Suppose that X = {x1 , . . . , xn } and Y = {y1 , . . . , ym }. Without loss of generality,
assume n ≤ m. Then there are r1 , . . . , rn ∈ D so that ym = r1 x1 + · · · + rn xn . Let k be the
smallest index with rk 6= 0. Then
xk = rk−1 ym − rk−1 rk+1 xk+1 − · · · − rk−1 rn xn .
So X1 = {x1 , . . . , xk−1 , xk+1 , . . . , xn } ∪ {ym } spans V . Now we do the same thing for ym−1
with X1 . Thus, we can find ai ∈ D and bm ∈ D so that
ym−1 = bm ym + a1 x1 + · · · + ak−1 xk−1 + ak+1 xk+1 + · · · + an xn .
If all ai = 0, then ym−1 = bm ym , but this is a contradiction as Y will no longer be linearly inde-
pendent. So there is ai so that ai 6= 0. Pick the smallest such index s so that as 6= 0. Using the
same argument as we did on xk , we see that xs ∈ hx1 , . . . , xk−1 , xk+1 , · · · , xs−1 , xs+1 , . . . , xn , ym ,
ym−1 i. Hence X \ {xs , xk } ∪ {ym , ym−1 } spans V . We can use this argument repeatedly (at
each step i, throw out xki from X, and add ym−i+1 ) till we reach step u = n − 1, where we
have
Xu = X \ {xk1 , . . . , xkn−1 } ∪ {ym , ym−1 , . . . , ym−u+1 }
spans V . Hence ym−u ∈ hXu i. This means we can throw out the last remaining xi (specifi-
cally, xku ), so Xu = {ym , . . . , ym−u } spans V . But this is possible only when Xu = Y . Hence
m − u = 1, or m = u + 1 = n − 1 + 1 = n, as required. 
Definition 10.2. We say that R a ring with unity has the invariant rank property if for
every free R-module F , any two bases have the same cardinality. In this case we call the
cardinality of a basis (of F ) the rank (or the dimension) of F .
Example. Any division ring has the invariant rank property. Any commutative ring has the
invariant rank property.

11. February 6
Our goal in this section is to prove that the rank of a free module is well-defined if it is a
module over a commutative ring with unity.
Lemma 11.1. Let R be a ring with unity, and I a proper ideal of R. Suppose that F is a
free R-module, X a basis of F , and Π : F → F/IF the canonical quotient map. Then F/IF
is a free R/I-module with basis Π(X) and |Π(X)| = |X|.
16
Proof. If y ∈ F/IF , then evidently there is x ∈ F such that y = x + IF . Let r1 , . . . , rn ∈ R
satisfy x = r1 x1 + · · · + rn xn . (note that r1 , . . . , rn , x1 , . . . , xn are unique by the linear
independence of a basis). Thus Π(x) = y = r1 (x1 + IF ) + · · · + rn (xn + IF ) = r1 Π(x1 ) +
· · · + rn Π(xn ). This means Π(X) spans F/IF .
Let r1 Π(x1 ) + · · · + rn Π(xn ) = 0 for some ri ∈ R and xi ∈ X (where ri := ri + I).
If Π(r1 x1 + · · · + rn xn ) = 0, then r1 x1 + · · · + rn xn ∈ IF . Then we know there exist
y1 , . . . , ym ∈ X and s1 , . . . , sm ∈ I such that
r1 x1 + · · · + rn xn = s1 y1 + · · · + sm ym .
Then by the uniqueness of presentation of an element of F in terms of X, we have m = n
and ri = si ∈ I, and yi = xi . So r1 , . . . , rn ∈ I, or r1 = · · · = rn = 0. Hence Π(X) is linearly
independent over R/I, meaning it is a basis of F/IF as an R/I-module.
As for the last part, we need to show that Π is one-to-one on X. If Π(x) = Π(x0 ), then
Π(x − x0 ) = 0. Thus x − x0 ∈ IF , so x − x0 = s1 y1 + · · · + sm ym for si ∈ I and yj ∈ X.
By the uniqueness of presentation, indeed m = 2; and without loss of generality we may let
y1 = x, y2 = x0 , s1 = 1, and s2 = −1. So 1 ∈ I, so I = R. But this contradicts the fact that
I is a proper ideal of R. Hence Π is one-to-one on X, from which |Π(X)| = |X| follows. 
Definition 11.1. If M is an R-module, then M has torsion if there exist non-zero r ∈ R
and m ∈ M such that rm = 0. M is said to be torsion-free if M has no torsion elements.
Proposition 11.1. Suppose R is an integral domain, and M an R-module. If M is free,
then M is torsion-free.
Proof (sketch). Suppose m is a torsion-element. Then there is r such that rm = 0. Then
there exist unique x1 , . . . , xn basis elements and r1 , r2 , . . . , rn ∈ R such that m = r1 x1 + · · · +
rn xn . So rm = rr1 x1 + · · · + rrn xn = 0. Thus rri = 0 for all i, so r = 0, which contradicts
the fact that r is non-zero. 
Remark 11.1. What happens if R is not an integral domain? Then there exist zero divisors
in R, i.e., r 6= 0, s 6= 0, but rs = 0. Suppose that F is a free R-module with basis X, and
x ∈ X. Since s 6= 0, indeed sx 6= 0. But r(sx) = (rs)x = 0x = 0, so we see that sx is a
torsion element. So a free module may contain a torsion element in this case.
Proposition 11.2. Suppose f : R → S is a surjective ring homomorphism (i.e., S is a
homomorphic image of R) and that both R and S contain identity. If S has the invariant
rank property, then R also has the invariant rank property.
Proof. If ker f =: I, then by the first isomorphism theorem, S ∼
= R/I. If F is a free R-
module, and X and Y are both bases of F , we want to show that |X| = |Y |. But this
follows from the first isomorphism theorem, Lemma 11.1, and the invariant rank property
of R/I ∼= S; therefore |X| = |Π(X)| = |Π(Y )| = |Y |. 
Theorem 11.1. Every commutative ring with unity has the invariant rank property.
Proof. R has a maximal ideal m by Zorn’s lemma, so R/m is a field, and we have a surjective
homomorphism R → R/m. So by Proposition 11.2, R has the invariant rank property.
Recall that R/m is a fortiori a division ring, so R/m has the invariant rank property. 
17
 12. February 8
12.1. Dimension theory in division rings
Theorem 12.1. Let D be a division ring, and V a vector space over D. Suppose that W is
a subspace of V . Then
(i) dimD W ≤ dimD V .
(ii) If dimD V < ∞ and dimD V = dimD W , then W = V .
(iii) dimD V = dimD W + dimD V /W .
Proof. (i) A basis X of W can be extended to a basis Y of V . So |X| ≤ |Y |, from which
dimD W ≤ dimD V follows.
(ii) Let X be a basis of W , and we proved X can be extended to a basis Y of V , so X ⊆ Y .
But then |X| = |Y | so X = Y . Therefore V = W .
(iii) Pick a basis X for W and extend to a basis Y for V . So X ⊆ Y . Let Z = {y +W : y ∈
Y \ X}. We want to claim that Z is a basis of V /W . Clearly Z ⊆ V /W , and if v + W ∈ V /W
then there exist unique y1 , . . . , yn ∈ Y and a1 , . . . , an ∈ D so that v = a1 y1 + · · · + an yn .
Then v + W = a1 y1 + · · · + an yn + W . Without loss of generality, suppose y1 , . . . , ys ∈ / X
but ys+1 , . . . , yn ∈ X. This implies v + W = a1 y1 + · · · + as ys + W ∈ hZi, so Z spans V /W .
We also need to prove linear independence. Suppose that a1 (y1 +W )+· · ·+an (yn +W ) = 0
fo some a1 , . . . , an ∈ D and y1 + W, . . . , yn + W ∈ Z. Suppose that there are b1 , . . . , bm ∈ D
and x1 , . . . , xm ∈ X such that a1 y1 + · · · + an yn = b1 x1 + · · · + bm xm . But since Y is linearly
independents, this forces ai = bj = 0 for all 1 ≤ i ≤ n, 1 ≤ j ≤ m. So Z is a basis of V /W .
Also |Z| = |Y | − |X| = dimD V − dimD W , from which the claim follows. 
Corollary 12.1. Let V and V 0 be D-modules, where D is a division ring. Let f : V → V 0
be a linear transformation (or, equivalently, a D-module homomorphism). Then there exists
a basis X of V such that X ∩ ker f is a basis of ker f , and f (X) \ {0} is a basis of im f .
Furthermore, dimD V = dimD ker f + dimD im f .
Proof. Apply the previous theorem (iii) with W = ker f which is a submodule of V . Recall
that any D-module is free since D is a division ring, so W has a basis X 0 which can be
extended to a basis X of V . Also, V /W ∼
= V / ker f ∼
= im f by virtue of the first isomorphism
theorem for modules. Therefore f (X) \ {0} is a basis of im f . 
Corollary 12.2. Let V and W be vector spaces over division ring D, and that both V and
W are finite-dimensional. Then dimD V + dimD W = dimD (V + W ) + dimD (V ∩ W ).
Proof. Exercise. 
13. February 11: Projective and injective modules
Definition 13.1. A module P over a ring R is said to be projective if given any diagram of
R-module homomorphisms whose bottom row is exact (i.e., g is an epimorphism),
P
h
f
g
A B 0
there exists an R-module homomorphism h : P → A that makes the above diagram commute
(gh = f ).
18
 Now we shall take a look at some examples of projective modules.
Theorem 13.1. Every free module F over a ring R with unity is projective.
Remark 13.1. The theorem holds even without the unity assumption.
Proof. Consider
F
h?
f
g
A B 0
with the bottom row exact. Let X be a basis of F . Let x ∈ X. Since g is an epimorphism,
there is ax ∈ A such that g(ax ) = f (x). Define h0 = x → A by h0 (x) = ax . Since F is free,
the map h0 induces an R-module homomorphism h : F → A defined by
n
! n
X X
h ci x i = ci axi .
i=1 i=1
P
Note that h is well-defined since F is free – F being free implies that ci xi is the unique
representation of an element of F . Now it is not a hard exercise to check that h is a
homomorphism. Now, we have f (x) = g(ax ) = gh(x). By the uniqueness of presentation
of elements of F (as F is free), we see that f (u) = gh(u) for all u ∈ F . Therefore F is
projective as required. 
Theorem 13.2. Let R be a ring with unity. The following conditions on an R-module P
are equivalent:
(i) P is projective.
f g
(ii) Every short exact sequence 0 −→ A −→ B −→ P −→ 0 is split exact. Hence
B∼= A ⊕ P.
(iii) P is a direct summand of a free module F . In other words, F ∼
= K ⊕ P with F a free
R-module and K an R-module.
Proof. ((i) ⇒ (ii)) Consider the diagram
P
h?
idP
g
B P 0
Since P is projective, there exists an R-module homomorphism h : P → B so that gh = idP .
Thus we have
f g
0 A B P 0
h
Therefore the above sequence splits, so B ∼
= A ⊕ P as required.
((ii) ⇒ (iii)) Every R-module is a homomorphic image of a free module. So there exists a
free module F such that
f
0 ker f B P 0
is exact. By hypothesis, the sequence splits so
F ∼= ker f ⊕ P.
19
Now take ker f =: K.
((iii) ⇒ (i)) Consider a diagram

F
ι π
h0
P
h
f
g
A B 0
with F ∼= K ⊕P . Since F is free, it is projective. So there exists an R-module homomorphism
h : F → A such that gh0 = f π. Define h : P → A as h = h0 ι. Then gh = gh0 ι = f πι =
0

f ◦ idP = f . 
Proposition
X13.1. Let R be a ring with unity, and let I be an index set . A direct sum of
R-modules Pi is projective if and only if each Pi is projective for all i ∈ I.
i∈I
P
Proof. (⇒) Suppose that Pi is projective. Then
X X
Pi = Pi ⊕ Pj
i∈I j∈I
| {z } j6=i
=:U | {z }
=:V

for a fixed i ∈ I. Now consider the diagram

U
ιi πi
h0
Pi
h
f
g
A B 0
Since U is projective, there exists an R-module homomorphism h0 : U → A such that
gh0 = f πi . Define h : Pi → A as h = h0 ιi . Then gh = gh0 ιi = f πi ιi = f idPi . So Pi is
projective for all i ∈ I.
(⇐) Suppose that Pi is projective for all i ∈ I. Consider the diagram

P
ιi πi
h0
U
h
f
g
A B 0
Since Pi is projective, there exists an R-module homomorphism h0i : Pi → A such that
gh0i = f ιi . By the universal property of direct sums, there exists an R-module homomorphism
20
h : U → A such that hιi = h0i . Then ghιi = gh0i = f ιi for all i ∈ I. Therefore gh = f as
needed. So X
U= Pi
i∈I
is projective. 

14. February 25 & 27

Definition 14.1. If R is a ring with identity, then an R-module J is called injective if for
any diagram of R-modules and R-module homomorphisms
J
f h
g
0 A B
there is h : B → J such that the diagram commutes, i.e., hg = f .
Lemma 14.1 (Baer’s criterion). Suppose R is a ring with the identity, and J an R-module.
Then J is injective if and only if for any left ideals I of R, any R-module homomorphism
I → J can be extended to an R-module homomorphism from R to J.
Proof. Let f : I → J and consider the diagram
g
0 I R
f
h
J
which is exact. Since J is injective, there is h : R → J such that hg = f .
(⇐) Suppose that we have the diagram of R-module homomorphisms
g
0 A B
f
∃?h
J
Consider the set S := {hC : C → J | im g ⊆ C ⊆ B}. We claim that S 6= ∅ since f g −1 :
im g → J is in S. S is partially ordered by ≤ where hc ≤ hD ⇔ C ⊆ D and hD |C = hC .
Suppose that C is a chain in S. We shall show that S has an upper bound in S. Write
[
MC := C.
hC ∈C

Then note that MC is a submodule of B containing im g. im g ⊆ MC ⊆ B, so we can define

the homomorphism hMC : MC → B defined by hMC (x) = hC (x) when x ∈ C and hC ∈ C.
Thus hMC ∈ S and is an upper bound for C. By Zorn’s lemma, S has a maximal element;
let this maximal element be M . So let hM : M → J.
g ⊆
0 A M B
f
hM
J ∃?h
21
So far, we know that there is hM making the above diagram commute. But is it M = B?
This is what we want. Suppose that M ( B. Then there is b ∈ B \ M . Construct
I = {r ∈ R : rb ∈ M }. This is an ideal (proving this is left as an exercise); consider now
f 0 : I → J defined by r 7→ hM (rb). f 0 is a well-defined R-module homomorphism (exercise
to prove that this is the case). Therefore, by assumption
ι
0 I B
f0
l
J
there is l : R → J such that lι = f 0 . Now define h : M +Rb → J where a+rb 7→ hM (a)+rl(1).
Suppose that a, a0 ∈ M and r, r0 ∈ R such that a + rb = a0 + r0 b. Then (r0 − r)b = a − a0 ∈ M .
Thus r−r0 ∈ I, so rl(1)−r0 l(1) = (r−r0 )l(1) = l((r−r0 )·1) = l(r−r0 ) = hM ((r−r0 )b). Hence
hM ((r − r0 )b) = hM (a0 − a) = hM (a0 ) − hM (a); it follows that hM (a) + rl(1) = hM (a0 ) + r0 l(1).
It is a straightforward verification to check whether h is an R-module homomorphism. This
means that h = hM +Rb ∈ S, which contradicts the maximality of hM . This forces M = B,
so hM = hB is indeed the homomorphism we were seeking. 
15. March 1
Definition 15.1. Let M be an R-module over domain R. If m ∈ M and r ∈ R, we say
that m is divisible by r if there is m0 ∈ M such that m = rm0 . We say that M is a divisible
module if every m ∈ M is divisible by every non-zero r ∈ R.
Example. Q is divisible Z-module. Frac(R), the fraction field of R, is a divisible R-module,
where R is a domain.
Proposition 15.1. If R is a domain, and M an injective R-module, then M is divisible.
Proof. Let m ∈ M and r ∈ R with r 6= 0; we need to find x ∈ M such that m = rx. Let
f : (r) = Rr → M so that f (ar) = am. f is well-defined since R is a domain, and f is an
R-module homomorphism. Since M is injective, by Baer’s criterion, there is h : R → M
such that h|(r) = f . Thus m = f (r) = h(r) = h(r · 1) = rh(1). Now let x = h(1), so we have
m = rx. The claim follows. 
Theorem 15.1. Suppose R is a principal ideal domain, and M an R-module. Then M is
injective if and only if M is divisible.
Proof. (⇐) Suppose that M is divisible. By Baer’s criterion, it suffices to show that for any
ideal I of R and any f : I → M an R-module homomorphism, f can be extended to the
entire R. Since R is a PID, there is a such that I = (a). Since M is divisible, there is
m ∈ M such that (a) = am ∈ M . Let h : R → M be h(r) = rm. One can verify that h
is an R-module homomorphism. If r ∈ I, then h(r) = rm; if s ∈ R satisfies r = sa, then
h(r) = rm = sam = sf (a) = f (sa) = f (r). Thus h extends f , so M is injective.
(⇒) This follows from Proposition 15.1. 

22
Corollary 15.1. Let R be a PID. Suppose M an injective (hence also divisible) R-module,
and N a submodule of M . Then M/N is injective (hence divisible) over R.
Proof. If m + N ∈ M/N and r 6= 0 ∈ R, then there exists m0 ∈ M such that m = rm0 .
Hence m + N = rm0 + N = r(m0 + N ). Therefore M/N is divisible. But then over a PID,
any module is divisible if and only if it is injective, so the claim follows. 
Corollary 15.2. The homomorphic image of a divisible group (i.e., divisible Z-module) is
divisible.
Proof. Let G0 be a homomorphic image of a divisible group G. So there exists a homomor-
phism ϕ : G → G0 such that ϕ is surjective. So by the first isomorphism theorem we have
G0 ∼
= G/ ker ϕ. G/ ker ϕ is divisible by the previous corollary, so G0 is also divisible. 

16. March 6 & 8

Recall that if M and N are R-modules, then HomR (M, N ) is the set of all R-module
homomorphisms from M to N .
Proposition 16.1. If J is a divisible abelian group, and R is a ring with identity, then
HomZ (R, J) is an injective R-module.
Proof. We know HomZ (R, J) is an R-module with action of R defined by rf (x) := f (xr),
where r ∈ R and f ∈ HomZ (R, J). Assume that I is a left ideal of R, and f : I → HomZ (R, J)
is an R-module homomorphism. We would like to apply Baer’s criterion: that is, find
ψ : R → HomZ (R, J) such that ψ extends f .
Let g : I → J be g(x) = f (x)(1). We need to verify if g is an R-module homomorphism.
Let x, y ∈ I and r ∈ R. Then g(rx + y) = f (rx + y)(1) = (rf (x) + f (y))(1) = rf (x)(1) +
f (y)(1) = rg(x) + g(y), as needed. So we have
0 I R
g
l
J
with 0 → I → R being an exact sequence. Since J is a divisible Z-module, so J is an
injective Z-module. Hence there exists l : R → J which is a Z-module homomorphism such
that l|I = g by Baer’s criterion. Now define h : R → HomZ (R, J) by r 7→ h(r) : R → J,
where h(r) maps x to l(xr).
(1) We need to verify if h(r) is a group homomorphism for any r ∈ R. For any x, y ∈ R
we have
h(r)(x + y) = l((x + y)r)
= l(xr + yr)
= l(xr) + l(yr) (because l is a group homomorphism)
= h(r)(x) + h(r)(y).
(2) h is well-defined. Let r = r0 where r, r0 ∈ R. Then for any x ∈ R we have h(r)(x) =
l(xr) and h(r0 )(x) = l(xr0 ). If r = r0 in R, then xr = xr0 in R, so l(xr) = l(xr0 ).
Hence h(r)(x) = h(r0 )(x), so h is well-defined.
23
 (3) h is an R-module homomorphism. Consider h(rx + y) : R → J. For any u ∈ R,
h(rx + y)(u) = l(u(rx + y)) = l(urx + uy)
= l(urx) + l(uy) (∵ l is a group homomorphism)
= h(x)(ur) + h(y)(u) = (rh(x))(u) + h(y)(u)
= (rh(x) + h(y))(u),
as required.
(4) Finally, we need h|I = f . Suppose r ∈ I. Then h(r) : R → J maps x 7→ l(xr). But
xr ∈ I since I is a left ideal. Therefore
l(xr) = g(xr) = f (xr)(1)
= xf (r)(1)
= f (r)(1 · x) (since f is an R-module homomorphism)
= f (r)(x).
Therefore for any r ∈ I, we have h(r)(x) = f (r)(x). Hence h = f whenever r ∈ I, so
h|I = f as desired. 
We want to prove that if R is a ring with identity and M an R-module, then M ⊆ J for
some injective R-module J.
First we want to prove this for the case R = Z.
Lemma 16.1. Every abelian group can be embedded in a divisible abelian group.
Proof. Let G be an abelian group. Then G is a Z-module, so there exists free Z-module F =
G. The first isomorphism theorem implies G ∼
L
Z and an epimorphism
L f : F →L = F/ ker f .
Observe that F = Z ,→ D = Q. D is divisible since Q is divisible as a Z-module.
Z is a PID, so L Q is injective as well as a Z-module; any direct sum of injective modules is
injective, so Q = D is injective as a Z-module.
If h is the injection from F to D, then F ∼= h(F ). Thus, G ∼= F/ ker f ∼
= h(F )/h(ker f ) ⊆
D/h(ker f ). So G is embedded in an injective Z-module; note that any quotient of a divisible
module is also divisible, making D/h(ker f ) divisible also. 
Theorem 16.1. Let R be a ring with identity, and M an R-module. Then M can be
embedded into an injective R-module.
Proof. Let M be an abelian group. By the previous lemma there exists a divisible group J
(injective Z-module) such that f : M ,→ J is a group monomorphism. We want to build
f : HomZ (R, M ) → HomZ (R, J) mapping g 7→ f g. Previously, we showed that HomZ (R, J)
is an injective R-module. We will show that M can be embedded here.
We claim that f is an R-module homomorphism. That is, if a ∈ R and g1 , g2 ∈ HomZ (R, M ),
then f (ag1 + g2 ) = f (ag1 + g2 ) = f (ag1 ) + f (g2 ) as f is a group homomorphism. Observe
that for any r ∈ R,
f (ag1 )(r) = f ((ag1 )(r)) = f (g1 (ra)) = f g1 (ra) = af g1 (r).
Therefore
f (ag1 + g2 ) = f (ag1 ) + f (g2 ) = af g1 + f g2 ,
as required.
24
 Now that we showed f is an R-module homomorphism, we now need to show that f is
injective. Suppose f (g) = 0. Then f g = 0, so in particular f g(1) = 0. Therefore f (g(1)) = 0;
but since f is injective, we have g(1) = 0. Thus g ≡ 0 as desired. Thus f is an R-module
monomorphism as needed, so HomR (R, M ) is a submodule of HomZ (R, M ).
Let ϕ : M → HomR (R, M ) be m 7→ fm where fm : R → M maps r to rm. Then ϕ is an
R-module monomorphism. Indeed, if ϕ(m) = 0, then fm (r) = 0 for all r ∈ R, which implies
fm (1) = 0. Therefore 1m = m = 0, as needed.
Now we have a chain of injections
ϕ i f
M ,→ HomR (R, M ) ,→ HomZ (R, M ) ,→ HomZ (R, J).
But then we previously proved that HomZ (R, J) is injective, so M is embedded in an injective
R-module as desired. 
Theorem 16.2. Let R be a ring with identity, and J an R-module. Then the following are
equivalent:
(i) J is injective.
(ii) Every short exact sequence 0 → J → B → C → 0 is split exact. In particular,
B∼ = J ⊕ C.
(iii) If J is a submodule of B, then J is a direct summand of B.
Proof. ((i) ⇒ (ii)) This works similarly to the projective case. Indeed,
f g
0 J B C 0
id
∃h
J
Since J is injective, there is h such that hf = idJ . By definition this is a split exact sequence,
so indeed B ∼ = J ⊕ C.
((ii) ⇒ (iii)) The exact sequence
0 −→ J −→ B −→ B/J −→ 0
is split exact by (ii), so B ∼= J ⊕ B/J.
((iii) ⇒ (i)) By the previous theorem, J ⊆ J 0 where J 0 is an injective R-module. By (iii)
J is a direct summand
Y of an injective module, so J is injective. Recall that a direct product
of R-modules Ji is injective if and only if Ji is injective for each i ∈ I. 
i∈I

17. March 11, 13 & 15

Recall that if A and B are R-modules then
HomR (A, B) = {f : A → B : f is a R-module homomorphism}.

Theorem 17.1. Let ϕ : C → A and ψ : B → D be R-module homomorphisms where R is a

ring. Then
θ : HomR (A, B) → HomR (C, D)
mapping f 7→ ψf ϕ is a group homomorphism.
25
 ϕ f ψ
Proof. Note that θ is well-defined since it is just a composition of functions (C → A → B →
D). θ is additive: for any f, g ∈ HomR (A, B), we have θ(f + g) = ψ(f + g)ϕ = ψf ϕ + ψgϕ =
θ(f ) + θ(g). 
Definition 17.1. We shall denote the θ in Theorem 17.1 by Hom(ϕ, ψ), and call it the
homomorphism induced by ϕ and ψ.
Note that ϕ1 : E → C, ϕ2 : C → A, ψ1 : B → D, ψ2 : D → F . Then
Hom(ϕ1 , ψ2 ) Hom(ϕ2 , ψ1 ) = Hom(ϕ2 ϕ1 , ψ2 ψ1 ).
Hom(ϕ2 ϕ1 ,ψ2 ψ1 )
HomR (A, B) HomR (E, F )

Hom(ϕ2 ,ψ1 ) Hom(ϕ1 ,ψ2 )

HomR (C, D)

Proposition 17.1. The following are equivalent:

ϕ ψ
(a) 0 −→ A −→ B −→ C is an exact sequence of R-modules.
ϕ ψ
(b) For every R-module D, 0 −→ HomR (D, A) −→ HomR (D, B) −→ HomR (D, C) is an
exact sequence of abelian groups, where ϕ : f 7→ ϕf and ψ : g 7→ ψg.
Proof. (⇐) Suppose D = ker ϕ, and suppose ι : D ,→ A be the inclusion map. Note that
ι ∈ HomR (D, A). ϕ(ι) = ϕι = 0: if x ∈ D = ker ϕ, then ϕ(ιx) = ϕ(x) = 0. Thus ι ∈ ker ϕ;
but since ϕ is injective by exactness, we have ι = 0. Hence D = ker ϕ = 0, so ϕ is injective.
Now pick D = A. Then im ϕ = ker ψ. So ψϕ(idA ) = 0. So ψϕ idA = 0, hence ψϕ = 0.
Therefore im ϕ ⊆ ker ψ.
For the other inclusion, we shall pick D = ker ψ, and let ι : D ,→ B. Indeed, ψ(ι) = ψι = 0.
Hence ι ∈ ker ψ = im ϕ. Thus there exists f ∈ HomR (ker ψ, A) so that ι = ϕ(f ). Hence
ι(x) = ϕ(f (x)) ∈ im ϕ, so ker ψ ⊆ im ϕ. So ker ψ = im ϕ as desired, thereby completing the
proof.
(⇒) Let D be an R-module. Suppose f ∈ ker ϕ. Then ϕ(f ) = 0. So ϕf = 0. Hence for
all d ∈ D we have ϕ(f (d)) = 0. But ϕ is injective, so f (d) = 0 for all d ∈ D which gives
f = 0. Therefore ϕ is injective.
We still need to prove that im ϕ = ker ψ. Let f ∈ im(ϕ). Then f = ϕ(g) for some
g ∈ HomR (D, A). Thus f (d) = ϕg(d) = ϕ(g(d)) ∈ im ϕ = ker ϕ. Hence ψ(f ) = 0 so
f ∈ ker ψ. Hence im ϕ ⊆ ker ψ. Conversely, let f ∈ ker ψ. Then ψ(f ) = ψf = 0. Therefore
for all d ∈ D we have ψf (d) = 0 = ψ(f (d)). Thus im f ⊆ ker ψ = im ϕ. ϕ is injective, so
ϕ : A → im ϕ is an isomorphism, by the first isomorphism theorem. Now we shall construct
f ϕ−1
h : D → im f ,→ im ϕ → A where f ∈ HomR (D, B). Then h ∈ HomR (D, A). Moreover,
f = ϕh = ϕ(h) by construction, so f ∈ im ϕ. Hence ker ψ ⊆ im ϕ, so indeed ker ψ = im ϕ,
as needed. 
We can prove the analogous result for HomR (·, D) using a similar reasoning.
ϕ ψ
Theorem 17.2. Let R be a ring. Then A → B → C → 0 is an exact sequence of R-modules
ψ ϕ
if and only if 0 → HomR (C, D) → HomR (B, D) → HomR (A, D) is an exact sequence of
Z-modules.
26
 In summary, HomR (D, ·) preserves left-exactness and the arrows; on the other hand,
HomR (·, D) flips arrows, and changes right-exactness to left-exactness.
Now we shall discuss some cases in which Hom is also right-exact.
Theorem 17.3. Let R be a ring. Then the following are equivalent.
ϕ ψ
(i) 0 → A → B → C → 0 is a split exact sequence of R-modules
ϕ ψ
(ii) 0 → HomR (D, A) → HomR (D, B) → HomR (D, C) → 0 is a split exact sequence of
Z-modules for every R-module D.
ψ ϕ
(iii) 0 → HomR (C, D) → HomR (B, D) → HomR (A, D) → 0 is a split exact sequence of
Z-modules for every R-module D.
Proof. ((i) ⇒ (iii)) 0 → A → B → C → 0 is split exact, so there are ψ1 : C → B such that
ψψ1 = idC . Consider ψ1 : HomR (B, D) → HomR (C, D) defined the usual way (f 7→ f ψ1 ).
Note that ψ1 ψf = ψ1 (ψf ) = ψ1 (f ψ) = f ψψ1 = f where f ∈ HomR (C, D). So the left-
exactness of HomR (·, D) gives us exactness everywhere but at ϕ.
Now we need to show that ϕ is surjective. We already know that there is ϕ1 : B → A such
that ϕ1 ϕ = idA . Let ϕ1 : HomR (A, D) → HomR (B, D) be the usual map, i.e., f 7→ f ϕ1 .
Observe that ϕϕ1 = idHomR (A,D) . Therefore ϕ is surjective. Indeed, if f ∈ HomR (A, D), then
ϕϕ1 (f ) = ϕ(ϕ1 (f )) = f , so f ∈ im ϕ.
The remaining directions are left as exercises. 
Theorem 17.4. Let R be a ring, and let P be an R-module. The following are equivalent.
(i) P is projective.
ϕ ϕ
(ii) If B → C → 0 is an exact sequence of R-modules, then HomR (P, B) → HomR (P, C) →
0 is an exact sequence of Z-modules.
ϕ ψ
(iii) If 0 → A → B → C → 0 is a short exact sequence of R-modules, then 0 →
ϕ ψ
HomR (P, A) → HomR (P, B) → HomR (P, C) → 0 is a short exact sequence of Z-
modules.
ϕ
Proof. ((i) ⇒ (ii)) Suppose B → C → 0 is exact, and let f ∈ HomR (P, C). Since P is
projective ,there is g ∈ HomR (P, B) such that ϕg = f .
P
g
f
ϕ
B C 0
Thus for any f there is g such that ϕ(g) = f , which shows that ϕ is surjective.
ϕ
((ii) ⇒ (i)) Consider an exact sequence B → C → 0 with surjective ϕ, and let f : P → C
be an R-module homomorphism. But since ϕ is surjective, there is g : P → B such that
ϕ(g) = f . Hence ϕg = f , so P is projective (see the commutative diagram above).
ϕ ψ
((ii) ⇒ (iii)) Suppose 0 → A → B → C → 0 is a short exact sequence. Then we know
ϕ ψ
0 → HomR (P, A) → HomR (P, B) → HomR (P, C) → 0
is exact for the first three arrows by the left exactness of Hom. The fourth arrow is also
straightforward due to (ii).
27
 ϕ
((iii) ⇒ (ii)) Given B → C → 0, we can build a short exact sequence 0 → ker ϕ → B →
C → 0. By (iii),
ϕ ψ
0 → HomR (P, A) → HomR (P, B) → HomR (P, C) → 0
ψ
is exact, so hence HomR (P, B) → HomR (P, C) → 0 is exact. 
The next theorem proves the injective counterpart.
Theorem 17.5. Let R be a ring, and let J be an R-module. The following are equivalent.
(i) J is injective.
ϕ ϕ
(ii) If 0 → A → B is an exact sequence of R-modules, then HomR (B, J) → HomR (A, J) →
0 is an exact sequence of Z-modules.
ϕ ψ
(iii) If 0 → A → B → C → 0 is a short exact sequence of R-modules, then 0 →
ψ ϕ
HomR (C, J) → HomR (B, J) → HomR (A, J) → 0 is a short exact sequence of Z-
modules.
Proof. Similar to the projective case. 

18. March 18 & 20

Definition 18.1. Let MR be a right R-module, and R N a left R-module, and let F be the
free Z-module on the set M × N . That is, F has a basis {e(m,n) : (m, n) ∈ M × N }. For
the simplicity of notation, write (m, n) := e(m,n) . Then the tensor product of M and N is
defined as the Z-module
M ⊗R N := F/Z,
where Z is the subgroup of F generated by the set
K := {(m + m0 , n) − (m, n) − (m0 , n), (m, n + n0 ) − (m, n) − (m, n0 ),
(mr, n) − (m, rn) | m, m0 ∈ M, n, n0 ∈ N, r ∈ R}
For any m ∈ M and n ∈ N , m ⊗ n := (m, n) + Z.
Proposition 18.1 (“The three rules”). Definition of tensor product implies the following
properties:
(i) (m + m0 ) ⊗ n = m ⊗ n + m0 ⊗ n
(ii) m ⊗ (n + n0 ) = m ⊗ n + m ⊗ n0
(iii) r(m ⊗ n) = mr ⊗ n = m ⊗ rn
Corollary 18.1. m ⊗ 0 = 0 ⊗ n = 0.
We shall see that M ⊗R N is a Z-module for any ring R. If R is commutative, we will
see that M ⊗R N is not just an abelian group, but is an R-module. We shall also see that
M ⊗R N is generated by {m ⊗ n : m ∈ M, n ∈ N }. Thus any typical element of M ⊗R N is
of the form
Xh
mi ⊗ ni
i=1
where m1 , . . . , mh ∈ M, n1 , . . . , nh ∈ N , and h ∈ N.
28
Definition 18.2. Let MR and R N be right and left R-modules respectively, and let Q be
an abelian group. Then a function f : M × N → Q is said to be middle-linear if for all
m, m0 ∈ M, n, n0 ∈ N , and r ∈ R, f satisfies the following three conditions.
(i) f (m + m0 , n) = f (m, n) + f (m0 , n)
(ii) f (m, n + n0 ) = f (m, n) + f (m, n0 )
(iii) f (mr, n) = f (m, rn)
In particular, the middle-linear map ι : M × N → M ⊗R N defined by ι(m, n) = m ⊗ n is
said to be the canonical middle-linear map.
Proposition 18.2 (Universal property of tensor products). Let MR be a right R-module
and R N a left R-module; let Q be an abelian group. If f : M × N → Q is a middle-linear
map, then there exists a unique group homomorphism f : M ⊗R N : Q such that the diagram
below commutes.
M ⊗R N
ι
f
f
M ×N Q
i.e., f = f ι. Moreover, M ⊗R N is the unique abelian group with this property.
Proof. As before, let F be a free Z-module with on M × N , and let
K := h(m + m0 , n) − (m, n) − (m0 , n), (m, n + n0 ) − (m, n) − (m, n0 ),
(mr, n) − (m, rn) | m, m0 ∈ M, n, n0 ∈ N, r ∈ Ri.
Then M ⊗R N = F/K by definition. By the universal property of free modules, for the
function f : M × N → Q, there exists a unique abelian group homomorphism f 0 : F → Q
such that f 0 ι0 = f .
ι0
M ×N F
f
f0

Q
Now if m, m0 ∈ M, n, n0 ∈ N and r ∈ R, we have f 0 ((m + m0 , n) − (m, n) − (m0 , n)) = 0.
Similarly, f 0 (α) = 0 for all α ∈ K. Hence K ⊆ ker f 0 . Therefore f 0 induces an abelian group
homomorphism f : F/K → Q such that f (m ⊗ n) = f 0 ((m, n)) = f (m, n).
Suppose that g is another group homomorphism g : M ⊗R N → Q such that gι = f . Then
for any (m, n) ∈ M × N , g(m ⊗ n) = gι(m, n) = f (m, n) = f ι(m, n) = f (m ⊗ n). Hence
g = f , which proves the uniqueness of f . Finally, the uniqueness of M ⊗R N comes from the
uniqueness of universal objects in categories. 
Definition 18.3. Suppose that R is a commutative ring, and A, B, C R-modules (note that
since R is commutative, every module is is both a left R-module and a right R-module). A
bilinear map f : A × B → C is a function satisfying the following three conditions for all
a, a0 ∈ A, b, b0 ∈ B, r ∈ R.
(i) f (a + a0 , b) = f (a, b) + f (a0 , b)
(ii) f (a, b + b0 ) = f (a, b) + f (a, b0 )
(iii) f (ra, b) = rf (a, b) = f (a, rb)
29
Remark 18.1. The (iii) from the above definition gives us the R-module structure on M ⊗R N
when R is commutative.
Remark 18.2. When A and B are R-modules for a commutative ring R, then A ⊗R B is an
R-module, and the canonical middle-linear map ι : A × B → A ⊗R B is in fact bilinear.
Recall that if R is a commutative ring, then M, N are left R-modules, then M ⊗R N is
a left R-module with action on R defined as r(m ⊗ n) = rm ⊗ n = mr ⊗ n = m ⊗ rn for
r ∈ R, m ∈ M, n ∈ N .
Example. We claim that Z/2Z ⊗Z Z/3Z = 0. Indeed, suppose that a = 3a ∈ Z/2Z and
b ∈ Z/3Z. Then a ⊗ b = 3a ⊗ b = 3(a ⊗ b) = a ⊗ 3b = a ⊗ 0 = a ⊗ 0 = 0.
The above example shows that the value of x ⊗ y depends very much on where x and y
live. We present another example which illustrates this point.
Example. We will see what 2⊗1 is in Z⊗Z Z/2Z. We have 2⊗1 = 2(1⊗1) = 1⊗2 = 1⊗0 = 0.
But on the other hand, in 2Z ⊗Z Z/2Z, we have 2 ⊗ 1 6= 0.
Proposition 18.3. Let R be a commutative ring, and let M, M 0 , N, N 0 R-modules. Suppose
that f : M → M 0 and g : N → N 0 are R-module homomorphisms. Then there exists a unique
R-module homomorphism f ⊗ g : M ⊗R N → M 0 ⊗R N 0 where (f ⊗ g)(m ⊗ n) := f (m) ⊗ g(n).
Proof. Define h : M × N → M 0 ⊗R N 0 by h(m, n) = f (m) ⊕ g(n). We need to show that h
is well-defined, but this is straightforward since f and g are. We also need to show that h is
bilinear. Let m, m0 ∈ M, n, n0 ∈ N , and r ∈ R.
h(m + m0 , n) = f (m + m0 ) ⊗ g(n) = (f (m) + f (m0 )) ⊗ g(n)
= f (m) ⊗ g(n) + f (m0 ) ⊗ g(n) = h(m, n) + h(m0 , n)
h(m, n + n0 ) = f (m) ⊗ g(n + n0 ) = f (m) ⊗ (g(n) + g(n)0 )
= f (m) ⊗ g(n) + f (m) ⊗ g(n0 ) = h(m, n) + h(m, n0 )
h(rm, n) = f (rm) ⊗ g(n) = rf (m) ⊗ g(n) = r(f (m) ⊗ g(n)) = rh(m, n)
h(m, rn) = f (m) ⊗ g(rn) = f (m) ⊗ rg(n) = r(f (m) ⊗ g(n)) = rh(m, n).
Hence h is bilinear map from M × N to M 0 ⊗ N 0 . By the universality of tensor products, h
extends to unique R-module homomorphism. 
f
Proposition 18.4 (Right-exactness of tensor). Suppose R is a commutative ring. Let M −→
g
N −→ K −→ 0 be an exact sequence of left R-modules. If D is any right R-module, then
D id ⊗f D id ⊗g
D ⊗R M −→ D ⊗R N −→ D ⊗R K −→ 0
is also an exact sequence of R-modules.
Proof. We will prove it the direct way. First, we claim that idD ⊗g is surjective. Note that
D ⊗R K is generated by elements of the form d ⊗ k, where d ∈ D and k ∈ K. Since
g is surjective, there exists n ∈ N such that g(n) = k. Hence d ⊗ k = (idD ⊗g)(d ⊗
n). Second, we need im(idD ⊗f ) = ker(idD ⊗g). im(idD ⊗f ) is generated by d ⊗ n where
d ∈ D and n ∈ im f = ker g. Thus (idD ⊗g)(d ⊗ n) = d ⊗ g(n) = d ⊗ 0 = 0. Hence
d ⊗ n ∈ ker(idD ⊗g). To prove the reverse inclusion, consider the canonical quotient map
30
π : D ⊗R N → D ⊗R N/ im(idD ⊗f ). Since im(idD ⊗f ) ⊆ ker(idD ⊗g), there is a unique
R-module homomorphism
ϕ : (D ⊗R N )/ im(idD ⊗f ) → D ⊗R K.
We show that ϕ is an isomorphism, which will show that ker(idD ⊗g) = im(idD ⊗f ). To do
this we shall show that ϕ has an inverse, by showing that there is a bilinear map ψ : D×K →
(D ⊗ N )/ im(idD ⊗f ) defined by (d, k) 7→ d ⊗ n + im(1D ⊗ f ) where n ∈ N is such that
g(n) = k. We show that ψ is well-defined bilinear map. Suppose that n, n0 ∈ N such that
g(n) = g(n0 ) = k. Then ψ(d, k) = d ⊗ n + im(idD ⊗f ) but also ψ(d, k) = d ⊗ n0 + im(idD ⊗f ).
Observe that d ⊗ n − d ⊗ n0 = d ⊗ (n − n0 ) ∈ im(idD ⊗f ). But then g(n) = g(n0 ) = k,
so g(n − n0 ) = 0. Thus n − n0 ∈ ker g = im f , so ψ is well-defined. Proving bilinearity is
straightforward, so this will be left as an exercise. So by the universality of tensor, there
exists ψ : D ⊗R K → (D ⊗R N )/ im(idD ⊗f ). Finally, observe ψψ = ψψ = id, thereby
proving that ψ is an isomorphism as desired. 
Remark 18.3. The above statement can also be proved using the exactness of Hom and the
observation that Hom(M ⊗R N, P ) ∼
= Hom(M, Hom(N, P )).

19. March 25
Definition 19.1. A functor F is a function from a caterogy to another category preserving
morphisms. F is covariant if F (f ) : F (A) → F (B) for f : A → B. F is contravariant if
F (f ) : B → A where f : A → B. F is exact if F takes short exact sequences to short exact
sequences.
Example. Let R be a commutative ring, and D an R-module. Then HomR (D, ·) is a covariant
functor which is exact if and only if D is projective. Similarly, HomR (·, D) is a contravariant
functor which is exact if and only if D is injective. The functor · ⊗R D is a covariant functor
which is exact if and only if D is a flat module.
Corollary 19.1. Let R be a commutative ring, and M, M 0 , N, N 0 all left R-modules. Also, let
f : M → M 0 and g : N → N 0 surjective homomorphisms. Then f ⊗ g : M ⊗R N → M 0 ⊗R N 0
is a surjective homomorphism of R-modules.
Proof. Applying the functor M ⊗R ·, we see that
id ⊗g
M
M ⊗R N −→ M ⊗R N 0 −→ 0
is exact. Similarly, we can apply the functor · ⊗R N 0 gives
f ⊗id 0
M ⊗R N 0 −→N M 0 ⊗R N 0 −→ 0
is exact. Note that if m ∈ M and n ∈ N , then (f ⊗ g)(m ⊗ n) = f (m) ⊗ g(n) = (f ⊗
idN 0 )(m ⊗ g(n)). Therefore f ⊗ g = (f ⊗ idN 0 ) ◦ (idM ⊗g) : M ⊗ N → M 0 ⊗ N 0 . Hence f ⊗ g
is surjective since other two are. 
Theorem 19.1. Let R be a commutative ring with unity. Suppose that A is a right R-module
and B a left R-module. Then A ⊗R R ∼
= A and R ⊗R B ∼ = B.
Proof. Define f : R × B → B by f (r, b) = rb. We show that f is bilinear.
f (r + r0 , b) = (r + r0 )b = rb + r0 b = f (r, b) + f (r0 , b)
31
 f (r, b + b0 ) = r(b + b0 ) = rb + rb0 = f (r, b) + f (r, b0 )
f (sr, b) = (sr)b = s(rb) = sf (r, b) = (rs)b = r(sb) = f (r, sb).
By the universal property of tensor product, there is a R-module homomorphism f : R ⊗R
B → B defined by r ⊗ b 7→ rb. We just need to show that f is bijective. f is surjective since
for any b ∈ B, we have b = 1 · b = f (1 ⊗ b). As for injectivity, suppose that
n
!
X
f ri ⊗ bi = 0
i=1

where r1 , . . . , rn ∈ R and b1 , . . . , bn ∈ B. Then

n
X
r i bi = 0
i=1

in B. Thus,
n n n n
!
X X X X
ri ⊗ bi = ri (1 ⊗ bi ) = (1 ⊗ ri bi ) = 1 ⊗ ri bi = 1 ⊗ 0 = 0.
i=1 i=1 i=1 i=1

Thus f is an R-module isomorphism as required. 

20. March 27: Modules over principal ideal domains

Definition 20.1. Let R be a ring, and M a left R-module. M is a Noetherian module if M
satisfies the ascending chain condition (ACC) of submodules, i.e., for any chain of submodules
M1 ⊆ M2 ⊆ · · · ⊆ Mk ⊆ Mk+1 ⊆ · · · , there exists N such that Mn = Mn+1 = · · · for all
n ≥ N . Therefore every ascending chain of submodules stabilizes. In particular, R is a
Noetherian ring if it satisfies the ascending chain condition on its ideals.
Theorem 20.1. If R is a ring, and M a left R-module, then the following are equivalent.
(1) M is Noetherian.
(2) Every non-empty set of submodules of M contains a maximal element under inclu-
sion.
(3) Every submodule of M is finitely generated.
Proof. ((1) ⇒ (2)) Suppose that M is Noetherian, and Σ a non-empty set of submodules
of M . Let M1 ∈ Σ, and suppose that M1 is not maximal. Then there exists M2 ∈ Σ with
M1 ⊆ M2 . If M2 is not maximal, there exists M3 such that M1 ⊆ M2 ⊆ M3 . Repeating this
step, we can build an ascending chain of modules in Σ. But since M is Noetherian, there
must exists N such that Mn = Mn+1 = · · · for all n ≥ N . Then Mk is a maximal element
of Σ.
((2) ⇒ (3)) Let N be a submodule of M , and we want to show that N is finitely generated.
Let
Σ = {N 0 | N 0 finitely generated submodule of N }.
Clearly 0 ∈ Σ so Σ 6= ∅. Now let N 0 be a maximal element of Σ. If N = N 0 , we are done. If
not, then N 0 ( N . So there exists x ∈ N but x ∈ N 0 . But then N 0 = hf1 , . . . , fs i for some
f1 , . . . , fs ∈ N since N 0 is finitely generated. Define N 00 = hf1 , . . . , fs , xi. But N 00 ) N , and
32
clearly N 00 ∈ Σ. But this contradicts the maximality of N 0 . Therefore N = N 0 , so N is
finitely generated.
((3) ⇒ (1)) Suppose that M1 ⊆ M2 ⊆ · · · is an ascending chain of submodules of M . Let
[
N := Mi ,
i≥1

so N is a submodule of M . Thus N is finitely generated, say N = hf1 , . . . , fs i for f1 , . . . , fs ∈

M . Thus there exists Ma1 , . . . , Mas such that f1 ∈ Ma1 , . . . , fs ∈ Mas . Without loss of
generality suppose that a1 ≤ a2 ≤ · · · ≤ as . Thus Ma1 ⊆ Ma2 ⊆ · · · ⊆ Mas ; note that
f1 , . . . , fs ∈ Mas , so Mas = N . Therefore we have Mn = Mn+1 for any n ≥ as , which is
precisely the ascending chain condition we wanted to show. 
Example. Any PIDs are Noetherian rings since every ideal is generated by one element.
Definition 20.2. If R is a domain, and M an R-module, then
tor(M ) = {x ∈ M | rx = 0 for some r ∈ R \ {0}}
is called the torsion submodule.
Remark 20.1. The emphasis on the word “the” in the above definition is intended, to em-
phasize that tor(M ) is the unique maximal torsion submodule of M . Observe that any
submodule of tor(M ) is also a torsion module.
Remark 20.2. If M is a free R-module, then tor(M ) = 0. Thus any free module is torsion-
free.
Definition 20.3. The annihilator of M is
ann(M ) = {r ∈ R : rn = 0 for all n ∈ M }.
Remark 20.3. Note that the following properties hold for ann(M ):
(1) If N is not a torsion submodule of M , then ann(N ) = (0).
(2) If N ⊆ L both submodules of M , then ann(L) ⊆ ann(N ), since if rL = 0 then
rN = 0.
(3) If, in addition to (2), R is a PID, then ann(L) = (a) ⊆ (b) = ann(N ), and so b | a. In
particular, if x ∈ M then ann(x) = (a) ⊇ ann(M ) = (b), so a | b.
(4) ann(M ) is an ideal of R. Indeed, 0 ∈ ann(M ), so ann(M ) is non-empty. If a, b ∈
ann(M ), then (a − b)x = ax − bx = 0 − 0 = 0 for any x ∈ M , so a − b ∈ ann(M ).
Finally, for any a ∈ ann(M ) and r ∈ R, we have (ra)x = r(ax) = r0 = 0 for any
x ∈ M . Hence ra ∈ ann(M ).

21. March 29
Theorem 21.1. Let R be a PID, and M a free R-module of rank n < ∞. Suppose that N
is a submodule of M . Then
(1) N is free of rank m where m ≤ n.
(2) There is a basis y1 , . . . , yn of M such that a1 y1 , . . . , am ym is a basis of N where
a1 , . . . , am ∈ R are such that a1 | a2 | · · · | am .
33
Proof. The claims hold trivially for N = 0, so assume that N 6= 0. Thus for all ϕ ∈
HomR (M, R), ϕ(N ) is an ideal of R; and since R is a PID, we have ϕ(N ) = (aϕ ) where
aϕ ∈ R. Define
Σ = {(aϕ ) | ϕ ∈ HomR (N, R)}.
Clearly 0 ∈ Σ so Σ is non-empty. Since R is Noetherian and Σ 6= ∅, Σ has a maximal
element, say (aν ) for some ν ∈ HomR (N, R). Therefore ν(N ) = (aν ) ⊃ (aϕ ) = ϕ(N ) for all
ϕ ∈ HomR (M, R). Let a1 := aν .
First, we prove that a1 6= 0. Let M be a free module P with basis, say, x1 , . . . , xn , and
projection homomorphisms πi : M → R defined by cj xj 7→ ci . Since N 6= 0, πi (N ) 6= 0
for some i. Hence there exists a non-zero element in Σ, which is enough to show that a1 6= 0,
since (a1 ) is a maximal element of Σ.
Second, we claim that if y ∈ N such that ν(y) = aν = a1 , then a1 | ϕ(y) for all ϕ ∈
HomR (M, R). Fix ϕ ∈ HomR (M, R) and let (ϕ(y), a1 ) = (d). Indeed, if ϕ(y) ∈ (d) and
a1 ∈ (d), then d | ϕ(y) and d | a1 . Conversely, if d ∈ (ϕ(y), a1 ) then d = r1 a1 + r2 ϕ(y) for
some r1 , r2 ∈ R.
Let ψ : r1 ν + r2 ϕ ∈ HomR (M, R). Then ψ(y) = r1 ν(y) + r2 ϕ(y) = r1 a1 + r2 ϕ(y). So
d ∈ ψ(N ); hence (d) ⊆ ψ(N ). Thus (a1 ) ⊆ (d) ⊆ ψ(N ) ⊆ (a1 ) since a1 is a maximal
element. Since (a1 ) = (d) = ϕ(N ), a1 | d and d | ϕ(y), so a1 | ϕ(y) as desired.
Let ϕ = πi be the projection onto the “i-th coordinate”. Then a1 | πi (y), which holds true
for every i. So there exists bi ∈ R such that πi (y) = bi a1 for each i = 1, 2, . . . , n. Suppose
that y1 = b1 x1 + · · · + bn xn . Then a1 y1 = a1 b1 x1 + · · · + a1 bn xn = π1 (y)x1 + · · · + πn (y)xn =
y. Thus a1 = ν(y) = ν(a1 y1 ) = a1 ν(y1 ). But since a1 6= 0, it follows ν(y1 ) = 1.
We claim that y1 can be a basis element of M , and a1 y1 can be a basis elements of N . Note
that it suffices to show instead that (a) M = Ry1 ⊕ ker ν and (b) N = Ra1 y1 ⊕ (N ∩ ker ν)
– observe that the main claim follows from (a) and (b) by extending {y1 } and {a1 y1 } to a
basis.
We prove (a) first. Suppose that x ∈ M . Then x = ν(x)y1 +(x−ν(x)y1 ) = ν(x−ν(x)y1 ) =
ν(x) − ν(x)ν(y1 ) = ν(x) − ν(x) · 1 = 0. So x − ν(x)y1 ∈ ker ν. Hence M = Ry1 + ker ν.
Now suppose that Ry1 ∩ ker ν is non-trivial. Then there is r ∈ R such that ry1 ∈ ker ν.
Since ν(ry1 ) = rν(y1 ) = 0, it follows r = 0 since ν(y1 ) = 1. Hence Ry1 ∩ ker ν is trivial, as
required.
As for (b), we start by assuming that x0 ∈ N so that ν(x0 ) ∈ (a1 ) = ν(N ). Then
a1 | ν(x0 ). Thus there exists b ∈ R such that ν(x0 ) = ba1 . Now consider the decomposition
x0 = ν(x0 )y1 + (x0 − ν(x0 )y1 ). Clearly ν(x0 )y1 = ba1 y1 ∈ Ra1 y1 . Observe that
ν(x0 − ν(x0 )y1 ) = ν(x0 ) − ν(x0 )ν(y1 ) = ν(x0 ) − ν(x0 ) = 0,
so x0 − ν(x0 )y1 ∈ ker ν ∩ N . Using the similar argument as used in part (a), we see that
Ra1 y1 ∩ (ker ν ∩ N ) = 0, so N = Ra1 y1 ⊕ (N ∩ ker ν).
Now that all the ground work is complete, we shall go back to prove the two statements
of the theorem. For (1), we will prove by induction on m, where m is the maximum number
of linearly independent elements of N . If m = 0, then N is a torsion module, but this in
turn implies N = 0. Indeed, since M is free over a PID, M is torsion-free, which in turn
implies that the only torsion element of M (hence of N ) is 0. If m > 0, then N ∩ ker ν has
the maximum m − 1 linearly independent elements. By induction hypothesis, N ∩ ker ν is
of rank m − 1. Therefore N is free of rank m, completing the proof of (1).
34
 The proof of (2) is also by induction, this time on n = rank(M ). ker ν is indeed a
submodule of M by (1), and ker ν is free. By part (a), rank(ker ν) = n − 1. So by induction
hypothesis applied to ker ν and its submodule N ∩ ker ν, there exists a basis {y2 , . . . , yn } of
ker ν such that a2 y2 , . . . , am ym is a basis of N ∩ ker ν, and a2 | a3 | · · · | am . By (a) we see
that y1 , . . . , yn is a basis of M ; and by (b), a1 y1 , . . . , am ym is a basis of N . Now it remains
to show that a1 | a2 . Let ϕ ∈ HomR (M, R) be such that ϕ(y1 ) = ϕ(y2 ) = 1 but ϕ(yi ) = 0 for
all i > 2. So a1 = ϕ(a1 y1 ) ∈ ϕ(N ). Since (a1 ) ⊆ ϕ(N ) ∈ Σ and (a1 ) is maximal in Σ, we
have ϕ(N ) = (a1 ). Similarly, a2 = ϕ(a2 y2 ) ∈ ϕ(N ), so a2 ∈ (a1 ), which proves a1 | a2 . 

22. April 1
Definition 22.1. An R-module M is cyclic if M = hxi for some x ∈ M .
Let π : R → M = hxi such that π(1) = x and hence π(r) = rx. Then π is surjective, so by
the first isomorphism theorem we have M ∼ = R/ ker π. But if R is a PID, then there exists
a ∈ R such that ker π = (a). Thus M ∼ = R/(a). Therefore, a cyclic module over a PID R is
of this form. Particularly, (a) = ann(M ).
Theorem 22.1 (Fundamental theorem of finitely generated modules over a PID). Suppose
R is a PID, and M is a finitely generated R-module. Then the following are true.
(1) M is isomorphic to the direct sum of finitely many cyclic modules. That is, there
exist r ∈ N ∪ {0} and non-units a1 , . . . , am ∈ R∗ such that a1 | a2 | · · · | am such that
M∼ = Rr ⊕ R/(a1 ) ⊕ · · · ⊕ R/(am ).
(2) From the above isomorphism, R/(a1 ) ⊕ · · · ⊕ R/(am ) is isomorphic to the torsion
submodule of M . In particular, M is a torsion R-module if and only if r = 0, and in
this case ann(M ) = (am ).
(3) M is torsion-free if and only if M is free.
Proof. (1) M is finitely generated, so let {x1 , . . . , xn } be a generating set for M of minimal
cardinality. Let Rn be the free R-module of rank n with basis b1 , . . . , bn . Define π : Rn → M
by r(bi ) = xi , and extend by R-linearity to Rn . But π is surjective, so the first isomorphism
theorem implies M ∼ = Rn / ker π. ker(π) is a submodule of M , and M is free over R which is
a PID, so ker(π) is free over R. Hence there exist a basis y1 , . . . , yn of Rn and a1 , . . . , am ∈ R
such that a1 | a2 | · · · | am and a1 y1 , . . . , am ym is a basis of ker(π) by virtue of Theorem 21.1.
Thus we have
Ry1 ⊕ Ry2 ⊕ · · · ⊕ Ryn
M∼ = Rn / ker π = .
Ra1 y1 ⊕ · · · ⊕ Ram ym
Define ϕ : Ry1 ⊕ · · · ⊕ Ryn → R/(a1 ) ⊕ · · · ⊕ R(am ) ⊕ Rn−m by ϕ(u1 y1 , . . . , un yn ) = (u1 mod
(a1 ), · · · , um mod (am ), um+1 , . . . , ur ). And so ker ϕ = Ra1 y1 ⊕ Ra2y2 ⊕ · · · ⊕ Ram ym ⊕ 0n−m .
Putting the isomorphisms together, we see
Ry1 ⊕ · · · Ryn
M∼= ∼
= R/(a1 ) ⊕ · · · ⊕ R/(am ) ⊕ Rn−m .
Ra1 y1 ⊕ · · · ⊕ Ram ym
If any of the ai is a unit, then R/(ai ) = 0 so we can drop that component from the direct
sum. This means we can assume that any of the ai ’s are non-units.
(2) This follows immediately, since ann(R/(ai )) = (ai ).
(3) Each R/(ai ) is a torsion R-module, so R is torsion-free if and only if M ∼
= Rr . 
35
Definition 22.2. Suppose R is a PID, and M a finitely generated R-module. Then there
are r ∈ N ∪ {0} and a1 | a2 | · · · | am non-units such that
M∼ = Rr ⊕ R/(a1 ) ⊕ · · · ⊕ R/(am ).
Then r is called the free rank or the Betti number of M . a1 , . . . , am are called the invariant
factors of M , unique up to multiplication by units. Finally, we call such presentation the
invariant factor form.
Remark 22.1. The r and the ai from the above definition are all unique, though this is yet
to be proved.
Any PID is a UFD, so R has unique factorization. So if a ∈ R, then a = upα1 1 · · · pαs s where
the pi ’s are primes, and u is a unit and αi > 0 for all 1 ≤ i ≤ s. And hence the ideals (pαi i )
α
are uniquely determined by a. It is also known that (pαi i ) + (pj j ) = R for any i 6= j since
α α
gcd(pαi i , pj j ) = 1 (i.e., (pαi i ) and (pj j ) are comaximal). By the Chinese remainder theorem,
R/(a) ∼= R/(pα1 ) ⊕ · · · ⊕ R/(pαs ).
1 s

Apply this to the invariant factor form of M to obtain the following theorem.
Theorem 22.2. If M is a finitely generated R-module over a PID R, then M is the direct
sum of finitely many cyclic R-modules whose annihilators are either (0) or generated by
powers of primes in R, i.e.,
M∼
= Rr ⊕ R/(pα1 ) ⊕ · · · ⊕ R/(pαt ),
1 t

where r ≥ 0, pα1 1 , . . . , pαt t are powers of not necessarily distinct primes p1 , . . . , pt ∈ R.

Definition 22.3. The pα1 1 , . . . , pαt t in the above decomposition are called the elementary
divisors of M , and the abote decomposition is called the elementary divisor form.

23. April 3
In this lecture we will prove the uniqueness of presentation of a finitely generated modules
over a PID (i.e., the uniqueness of the Betti number, invariant factors, and elementary
divisors).
Theorem 23.1 (Primary decomposition theorem). Let R be a PID, and M a non-zero
torsion R-module (not necessarily finitely generated) with a non-zero annihilator a. Suppose
that the factorization of a into distinct powers of primes in R is a = upα1 1 · · · pαnn where u
is a unit, pi primes, and ai ∈ Z+ . Also let Ni = {x ∈ M : pαi i x = 0} for each 1 ≤ i ≤ n.
Then Ni is a submodule of M with annihilator pαi i and is the submodule of M consisting of
all elements annihilated by some power of pi . We have
M = N1 ⊕ N2 ⊕ · · · ⊕ Nn .
If M is finitely generated, then each Ni is a direct sum of finitely many cyclic modules whose
annihilators are divisors of pαi i .
Proof. The result is known if M is finitely generated (just group together all factors R/(pα ),
with the same p and varying α). In general, it is easy to prove that Ni is a submodule with
α
annihilator (pαi i ). If R is a PID, then (pαi i ) and (pj j ) is comaximal if i 6= j. Therefore by the
Chinese remainder theorem it follows M = N1 ⊕ N2 ⊕ · · · ⊕ Nn . 
36
Lemma 23.1. Let R be a PID, p a prime in R, and let F = R/(p) which is a field. Then
(1) If M = Rr , then M/pM ∼ = F r.
(2) If M = R/(a) and a 6= 0, then
(
F (if p | a in R)
M/pM ∼ =
0 (if p - a in R).
(3) M = R/(a1 ) ⊕ · · · ⊕ R/(ak ) where p | ai for all i, then M/pM ∼
= F k.
Proof. (1) Consider the map π : Rr → F r = (R/(p))r defined by (α1 , . . . , αr ) 7→ (α1 , . . . , αr )
where αi = αi mod (p). π is a surjective R-module homomorphism and π(α1 , . . . , αr ) = 0
if and only if p | αi for all i = 1, 2, . . . , r. Therefore ker π = pRr = pR ⊕ · · · ⊕ pR. Hence
Rp /pRr ∼= Fr ∼= M/pM .
(2) Let M = R/(a). Then pM = pR/(a) = ((p) + (a))/(a). If d = gcd(p, a), then
(p) + (a) = (d). So putting the two things together, we have
R/(a)
M/pM ∼
= ∼
= R/((p) + (a)).
((p) + (a))/(a)
Therefore if p | a, then R/(p) = F . If p - a, then gcd(p, a) = d = 1 so (d) = R. Therefore in
this case M/pM = 0.
(3) If M = R/(a1 )⊕· · ·⊕R/(ak ) such that p | ai for all i, then let π : R/(a1 )⊕· · ·⊕R/(ak ) →
R/(p)⊕· · ·⊕R/(p) be (u1 +(a1 ), . . . , uk +(ak )) → (u1 +(p), . . . , uk +(p)) where u1 , . . . , uk ∈ R.
Note that (u1 +(a1 ), dots, uk +(ak )) ∈ ker π if and only if p | ui for each i; this is also equivalent
to saying that ui + (ai ) ∈ pR/(ai ). This means that
ker(π) = pR/(a1 ) ⊕ · · · ⊕ pR/(ak ) = pM.
Therefore M/pM = M/ ker π ∼
= F k. 

24. April 5
Definition 24.1. If R is a ring, and M an R-module, then the p-primary submodule of M
is the submodule of M consisting of elements annihilated by a power of p.
Theorem 24.1 (Fundamental theorem of finitely generated modules over a PID – unique-
ness). Two finitely generated modules M1 and M2 over a PID R are isomorphic if and only
if they have the same free rank and the same list of invariants. Also, two finitely generated
modules M1 and M2 over a PID R are isomorphic if and only if they have the same free rank
and the same set of elementary divisors.
Proof. (⇐) This direction is evident (for both invariant factors and elementary divisors).
(⇒) Suppose that M1 ∼ = M2 , with an isomorphism ϕ : M1 → M2 . Note that then
ϕ(tor(M1 )) = ϕ(tor(M2 )) since am1 = 0 if and only if aϕ(m1 ) = 0. Hence
R r1 ∼
= M1 / tor(M1 ) ∼
= M2 / tor(M2 ) ∼
= Rr2 .
So by the invariant rank property of free modules over a PID, we see r1 = r2 . Hence we may
assume that M1 and M2 are both torsion modules. Suppose p is a prime, α ∈ Z+ , and pα
an elementary divisor of M1 . Suppose that M1 → M2 is an isomorphism. Then there exists
m1 ∈ M1 such that pα m1 = 0, so pα ϕ(m1 ) = 0. Thus the p-primary submodule of M1 is
37
isomorphic to the p-primary submodule of M2 . Observe that the p-primary component of
M1 is a direct sum of R/(pα ) for various α, and the same goes for M2 .
So without loss of generality, we may assume that we have two modules M1 and M2 where
ann(M1 ) and ann(M2 ) are both generated by a power of p – say ann(M1 ) ∼ = ann(M2 ) = (pk ).
We will prove by induction on k that M1 and M2 have the same list of elementary divisors.
If k = 0, then M1 = M2 = 0, so this completes the base case. Suppose k > 0. The. In M1
and M2 have elementary divisors p, p, . . . , p, pα1 , . . . , pαs . In other words,
| {z }
m times

M1 ∼
= (R/(p)) ⊕ R/(pα1 ) ⊕ · · · ⊕ R/(pαs ),
m

where 2 ≤ α1 ≤ α2 ≤ · · · ≤ αs . Now the module pM has elementary divisors pα1 −1 , . . . , pαs −1 .

Therefore,
pM1 ∼= Rm ⊕ R/(pα1 −1 ) ⊕ · · · ⊕ R/(pαs −1 ).
Similarly, the elementary divisors of M2 are p, p, . . . , p, pβ1 , . . . , pβt where 2 ≤ β1 ≤ · · · ≤ βt ,
| {z }
n times
so the elementary divisors of pM2 are pβ1 −1 , . . . , pβt −1 .
If M1 ∼ = M2 , then pM1 ∼ = pM2 . Furthermore, ann(pM1 ) ∼
= ann(pM2 ) = (pk−1 ). By the
induction hypothesis, we have β1 − 1 = α1 − 1, . . . , βt−1 = αs − 1. Hence s = t and αi = βi
for all 1 ≤ i ≤ s.
Also, if F := R/(p), we have F t+m ∼ = M1 /pM1 ∼= M2 /pM2 ∼= F t+n by Lemma 23.1, so
t + m = t + n, or m = n. Hence M1 and M2 have the same set of elementary divisors
p, p, . . . , p, pα1 , . . . , pαt .
| {z }
m times
We shall now show that M1 and M2 have the same invariant factors. If a1 | a2 | · · · | am are
invariant factors of M1 and b1 | b2 | · · · | bn those of M2 , then we can find elementary divisors
of M1 by factoring a1 , . . . , am , and of M2 by factoring b1 , . . . , bn . Since a1 | · · · | am , am
contains the largest power of each prime appearing in a1 , . . . , am−1 . Similarly, am−1 contains
the largest power of each prime appearing in a1 , . . . , am−2 , and so forth.
In a similar fashion, we get elementary divisors of M2 from b1 , . . . , bn . Since the list of
elementary divisors of M1 and M2 are the same, am and bn can only differ by a unit (i.e.,
am = ubn for some unit u ∈ R). This hold for am−1 and bn−1 , and so on. Hence m = n and
ai = ui bi for all 1 ≤ i ≤ n where each ui is a unit. 
Corollary 24.1. Let R be a PID, and M a finitely generated R-module.
(1) The elementary divisors of M are the prime power factors of the invariant factors of
M.
(2) The largest invariant factor of M is the product of the largest of the distinct prime
powers amongst the elementary divisors of M ; the next largest invariant factor of M
is the product of the largest of the remaining distinct prime powers, and so forth.
Corollary 24.2 (Fundamental theorem of finitely generated abelian groups). If G is a
finitely generated abelian group, then
(1) there exist r, n1 , . . . , ns ∈ Z satisfying G ∼
= Zr ⊕ Z/n1 Z ⊕ · · · ⊕ Z/ns Z such that:
(a) r ≥ 0, nj ≥ 2 for all j
(b) n1 | n2 | · · · | ns .
(2) The expression in (1) is unique.
38
 25. April 8
Recall that if R is a commutative ring and D an R-module, then HomR (D, ·) is a covariant
left exact functor, whereas HomR (·, D) is a contravariant left exact functor. Also, · ⊗R D or
D ⊗R · is a covariant right exact functor.
This is where the Tor module and the Ext module come in. Note that all of the afore-
mentioned functors do not entirely preserve exactness; but adding Tor for the tensor functor
compensates for lack of exactness; the Ext module does this for the Hom functors.
Recall that HomR (D, ·) presnrves exactness if and only if D is projective; the similar claim
hold for HomR (·, D) where D is injective. D needs to be flat in order for D ⊗R · or · ⊗R D
to preserve exactness. Observe that every R-module M is the homomorphic image of a
projective (or even free) module. Say M is generated by the subset X. Let F0 be the free
R-module on the set {ιx : x ∈ X}. Let α0 : F0  M be a surjective homomorphism defined
by α0 (ιx ) = x. Then the sequence
0 α
0 −→ ker α0 −→ F0 −→ M −→ 0
is exact. Now construct F1 for ker α0 so that α1 : F1 → ker α0 is a surjective homomorphism.
α3 α2 α1 α0
· · · F3 / F2 / F1 / F0 / M / 0
< <
ι
" α01 "
ker
;
α1 ker
;
α0

# #
0 0 0

Definition 25.1. An exact sequence of the form

· · · Fi −→ Fi−1 −→ · · · −→ F0 −→ M −→ 0
is called a free resolution of M where each Fi is a free R-module. If each Fi is projective,
then this is called a projective resolution of M .
Example. Let M = k[x, y, z]/(xy, yz) where k is a field. View M as an R-module where
R = k[x, y, z]. Let {e1 , e2 } be a basis of R2 and let the rightmost R be F0 . Then

α2 α1 α0
0 R R2 R R/i 0
α01

ker α1 ker α0

0 0

Note that ker α1 = hze1 − xe2 i and ker α0 = I since α10 (ze1 − xe2 ) = z(xy) − x(yz) = 0.
Suppose that N is an R-module with free (or projective) resolution
1 α
0 α
· · · −→ F1 −→ F0 −→ N −→ 0.
39
Applying M ⊗R · to the resolution of N gives
γi+1 γi γ1
C : · · · → M ⊗R Fi+1 → M ⊗R Fi → · · · → M ⊗R F1 → M ⊗R F0 → M ⊗R N → 0.
C is a chain complex such that im γi+1 ⊆ ker γi for all i.
Definition 25.2. Tori (M, N ) is the i-th homology module Hi (C) = ker γi / im γi+1 .
Remark 25.1. Tori (M, N ) is independent of which resolution of N one takes. Also, Tori (M, N )
remains invariant regardless of whether one starts with a projective resolution of M or of N .
Finally, Tor is a derived functor in the following sense. If M is a left R-module, and
0→A→B→C→0
is a short exact sequence of right R-modules, then there exists a long exact sequence
Department of Mathematics and Statistics, Dalhousie University, 6316 Coburg Rd, Hal-
ifax, NS, Canada B3H 4R2
E-mail address: hsyang@dal.ca

40