M3P8 Algebra III

July 13, 2011

Life gives you Lemons.

- Professor Kevin Buzzard

Chapter 1

Basics Properties of Rings

Definition 1.1. A ring (with 1) is a set R along with elements 0, 1 R and
maps + : R R R, : R R R (write a + b for addition and abbreviate
a b by ab) such that
(1) (R, +) is an abelian group with 0 as identity
(2) (R, ) is a semigroup with identity 1 (i.e. a, b R, (ab)c = a(bc))
(3) a, b, c, a(b + c) = ab + ac and (a + b)c = ac + bc.
Example.
- Z, Q, R, C with the obvious choices for 0, 1, +, are rings
- R[x] (the polynomials with real coefficients) is a ring
Observe that R is a group under +. So a R, inverse b R such that
a + b = 0. We will call this inverse a: (a) + a = 0. Further, we define
x y = x + (y) as subtraction. In general we cannot do division in rings!
Note. Some people do not demand 1 R and do not demand 1r = r = r1. For
these people, I define a ring with 1. Other people demand 0 6= 1 as an extra
axiom. This barely makes any difference as 0 = 1 = R = {0}.
Example. Mn (R) along with usual +, and 0 = 0n is a ring. The identity
element is 1 = In (Note that (AB)C = A(BC) but AB 6= BA in general.
Definition 1.2. A ring R is commutative if ab = ba a, b R.
Example. Residue Class Rings. Take m 1 to be an integer and define an
equivalence relation on Z by
a b m divides a b a b mod m.
Let Z/mZ denote the set of equivalence classes:
Z/mZ = {0, 1, . . . , m 1} = {[0]m , [1]m , . . . , [m 1]m }.
2

It turns out that Z/mZ is a ring under the operations

[a] + [b] = [a + b]
[a] [b] = [ab]
E.g. Z/5Z is a ring.
Note. Our intuition is bases on rings like Z, Q, . . . which are all well-behaved
rings. In general rings are not so well-behaved!
Example.

- Let R = M2 (R) and a = b =

0
0

1
0


. Then a 6= 0, b 6= 0, but ab = 0.

- Let R = Z/6Z, = [2], b = [3]. Then a 6= 0, b 6= 0, but ab = [0].

- Polynomials can have too many roots, e.g. consider x2 1 in the ring
Z/8Z: x = [1], [3], [5], [7] are all roots of the polynomial.
- Cancellation can also fail, i.e. ra = rb ; a = b. E.g. let r = [2], a =
[2], b = [0] in Z/4Z.
All this happens because we cannot divide.
Definition 1.3. A ring R is called a division ring if R/{0} is a group under
multiplication with 1 as the identity (1 6= 0). In other words, a 6= 0, b such
that ab = ba = 1.
Example.
- Z is not a division ring (as 2 does not have an inverse under )
- Z, Q, R are all division rings under the usual operations of addition and
multiplication.
A commutative division ring is called a field. For example, Z, Q, R are fields.
Fields are great: the whole theory of vector spaces works over a general field.
Example. The ring of polynomials in R, R[x], is not a field, as
polynomial.

1
x

is not a

Definition 1.4. If R is a ring and S R is a subset, we say S is a subring if

0, 1 S and t + s, st S s, t S and furthermore if S becomes a ring itself
with this 0, 1, +, , i.e. S satisfies the axioms.
Lemma 1.5. If R is a ring and S R is a subset of R such that 0, 1 S and
s + t, st, s t S s, t S then S is a subring of R.
Proof. We need to check the three axioms: (S, +) is a group, because s, t
S = s t S. So inverses exist. Also it is obviously abelian and the other
axioms are obvious, e.g. say r, s S. Need r(s + t) = rs + rt : but this is true
in R, so it must be true in S.
3

Example.
- Z is a subring of Q
- Q is a subring of R
- R is a subring of C.

- Let d Z be an integer that is not a square. Define Z[ d] to be the

following subset of C :

Z[ d] = {a + b d|a, b Z}.

Lemma 1.6. Z[ d] is a ring.

Proof.It is a subset
of C so letsuse Lemma 1.5:
We need to check that
if
a + b d =r Z[ d] and a0 + b0 d = r0 Z[ d], then r r0 , rr0 Z[ d].
r r0 Z[ d] is easy to check and

rr0 = (a + b d)(a0 + b0 d) = (aa0 + bb0 d) + (ab0 + ba) d.

Z

A slightly less obvious fact about Z[ d]:if a + d d= a0 + d0 d, then a =

a0 , b = b0 (where a, a0 , b, b0 Z). For a + d d = a0 + d0 d,

a a0 =b0 d d d

= (b0 b) d.
If b 6= b0 , then
a = a0 .

d=

aa0
bb0

Q, but

d 6 Q. Hence b = b0 = a a0 = 0 =

Example. Q[ d] = {a + b d|a, b Q} (for d 6 Z) is a ring (by the same

proof as before: subring of C).

Lemma 1.7. Q[ d] is a field.

Proof. Q[
d] is clearly commutative,
so all I need to do
is to check

that if
0 6= r Q[ d], then 1r Q[ d]. So assume 0 6= r = a + b d Q[ d]. Then

2
a2 b2 d 6= 0 for if a2 b2 d = 0, then a2 = b2 d and either b = 0 or d = ab .
But d is not square by assumption, hence
b = 0 = a2 = 0 = a = 0 = r = 0
contradiction. So Q 3 t = a2 b2 d 6= 0 and from above we see that

a
b
t t d.

1
r

Example. The set of all functions f : [0, 1] R is naturally a ring. the role
of 0 is played by the function sending any x [0, 1] to 0 and 1 is the function
sending any x [0, 1] to 1. Define f + g by (f + g)(x) = f (x) + g(x) and f g by
(f g)(x) = f (x)g(x).

Exercise. This defines a ring.

Proposition 1.8. Let R be a ring and say r, s, ri , si R.
(a) r0 = 0r = 0 r, s R
(b) if r denotes the inverse of r under addition, then
(r)s = r(s) = (rs)

(r)(s) = rs r, s R

P
 P
Pn
n
n
(c) ( i=1 ri )
j=1 sj =
i,j=1 ri sj
(d) if r R and rs = s s R, then r = 1.
(e) if R is a ring and 0 = 1 in R then R = {0} has one element (conversely,
{0} is a ring).
Proof.
(a) standard exercise in group theory: 0 + 0 = 0, hence
r(0 + 0) = r(0) = r0 + r0 = r0 = r0 = 0. Similarly for 0r = 0.
(b) to check (r)s = (rs) is need to check that (r)s + rs = 0. Then by
distributivity, it suffices to prove that(r + r)s = 0. But r + r = 0 and
0s = 0 by (a). Hence r(s) = rs. Now
(r)(s) = (r(s)) = (rs) = rs
since R is an additive group.
(c) tedious induction on m + n using distributivity.
(d) set s = 1.
(e) if r R, then r = r1 = r0 = 0 by (a). Conversely, check that {0} satisfies
all the axioms.
Convention: By definition 0, 1 R and define 2 R to be 1 + 1. Similarly for
3, 4, ..., 73, .... Further, define 1 R to be the additive inverse of 1 such that
1 + (1) = 0 and similarly n to be the additive inverse of n. We obtain a map
Z R which may or may not be an injection, e.g. 73 = 0 in Z/73Z and 73 = 1
in Z/72Z.
Definition 1.9. If R is a ring and if 0 6= a R, then we say a is a left-divisor
of zero if b 6= 0 in R such that ab = 0 (similarly for right-divisor of zero. Note
that if R is commutative, these notions coincide and we say that a is a zero
divisor). If a R and b R such that ab = ba = 1, then we say a is a unit
(for R commutative, we only need ab = 1 for a to be a unit). Write R for the
set of units in R.
Remark. R is a group, as associativity and identity are ring axioms and inverses
exists by definition of a unit.

Example.
- 2 is a zero divisor in Z/6Z as 2 3 = 0 in this ring but 2 6= 0, 3 6= 0.
- 5 is a unit in Z/6Z since 5 5 = 1 in this ring.


0 1
- The matrix
is a left and ring zero divisor in M2 (R) as
0 0


0
0

1
0

2
=0

(note that if a R and an = 0 for some n 1 then a is nilpotent).

- the units in M2 (R) are the invertible matrices (i.e. GL2 (R)) .
- R = Z has no zero divisors as ab = 0 = a = 0 or b = 0.
- The units in Z are Z = {1}.
- If R is a field (or even a divison ring), then R = R \ {0}.
Definition 1.10. A ring R is an integral domain if
(1) R is commutative
(2) 0 6= 1
(3) R has no zero divisors (i.e. if ab = 0, then a = 0 or b = 0).
Example.
- Z is an integral domain
- any field is an integral domain
- the zero ring {0} is not an integral domain (which is a wise convention).
- any subring of a an integral
domain is again an integral domain = any

subring of C, e.g. Z[ d], Q[ d] etc., is an integral domain.

Lemma 1.11. Let m be a positive integer and let R be the ring Z/mZ. Then
R is an integral domain iff m is prime.
Proof. Note first that if m = 1 is not prime, then Z/1Z = {0} is not an integral
domain. If m = p is prime, then I need to check that Z/pZ is an integral domain:
clearly, we have that Z/pZ is commutative and 0 6= 1. Now say that a, b 6= 0 in
Z/pZ : lift a to A Z and b to B Z. Since a, b 6= 0 we have that p - A, p - B,
hence p - AB. So AB 6= 0 mod p = ab 6= 0. So Z/pZ is an integral domain
if p is prime. To show the converse, assume that m is not prime, i.e. m = ab
with 1 < a < b < m. Then a, b 6= 0 and ab = 0 in Z/mZ. So Z/mZ is not an
integral domain.
Lemma 1.12.

(i) if R is a ring and a R with ar = as, then r = s

(ii) if R is an integral domain and if a 6= 0 and ar = as, then r = s.
Proof. For the first part, choose b R such that ba = 1. Then
ar = as = bar = bas = 1r = 1s = r = s.
For the second part, let a 6= 0 and ar = as. Then a(r s) = 0. But R is an
integral domain and a 6= 0. So
r s = 0 = r = s.

Note. (ii) is not a special case of (i), for example 2 Z is non-zero but not a
unit.
Abstract Polynomial Rings
Let R be any commutative ring. Define the polynomial ring R[x] of polynomials to be, formally, the set of all infinite sequences (c0 , c1 , . . . , cn , . . .) with
ci R i but all but finitely many ci equal to zero. Informally, we think
of (c0 , c1 , . . . , cn , 0, 0 . . .) as being c0 + c1 x + . . . + xn . Define 0 = (0, 0, . . .),
1 = (1, 0, 0, . . .) and
(a0 , a1 , . . .) + (b0 , b1 , . . .) = (a0 + b0 , a1 + b1 , . . .)
and
(a0 , a1 , . . .)(b0 , b1 , . . .) = (c0 , c1 , . . .)
with cn =

Pn

i=0

ai bni.

Exercise. This defines a ring.

Notation. Call this ring R[x] and write f = (a0 , a1 , . . . , ad , 0, 0, . . .) = a0 +a1 x+
. . . + ad xd . If ad 6= 0 then we say that d is the degree of f if an = 0 n > d.
Proposition 1.13. If R is an integral domain then R[x] is also an integral
domain.
Proof. Say 0 6= a, b R[x] . Write
a = a0 + a1 x + . . . + ad xd
b = b0 + b1 x + . . . + be x e
with ad , be 6= 0. Then ab = STUFF+ad be xd+e . Now, as R is an integral domain,
ad be 6= 0 = ab 6= 0.
If R is a commutative ring, define R[x1 , x2 ] = (R[x1 ])[x2 ].
Corollary 1.14. If R is an integral domain, then so is R[x1 , . . . , xn ].

Proof. Do induction on n.
We say that a subfield of a ring R is a subring S R that is a a field. For
example, R is a subring of R[x] (the constant polynomials) but also a subfield
of R[x].
Remark. If K is a subfield of the ring R, then R is naturally a vector space over
K. For example, C is a vector space over R.
Lemma 1.15. A finite integral domain is a field.
Proof. Say that 0 6= a R with R being a finite integral domain. We need to
find an inverse for a, i.e. b such that ab = 1. Consider the map ma : R R
given by ma (r) = ar for r R. ma is injective, for if ma (r) = ma (s), then
ar = as = r = s by 1.12(ii). Hence ma is injective. Also ma is surjective
since R is finite. Hence it is a bijetion, so we can choose b such that ma (b) =
ab = 1.
Trickier: a finite division ring is a field. This is known as the Artin-Wedderburn
Theorem.
Corollary 1.16. The ring Z/mZ is a field iff m is prime.
Proof. By 1.11 and 1.15 and the fact that a field is an integral domain.

Chapter 2

Homomorphisms, Ideals &

Quotient Rings
Definition 2.1. Let R and S be rings. A map : R S is a ring homomorphism if
(i) (0) = 0, (1) = 1
(ii) (a + b) = (a) + (b)
(ii) (a b) = (a) (b)
Remark. From Group Theory, we know (x) = (x) and so (x y) =
(x) + (y)
Definition 2.2. : R S is an isomorphism if a ring homomorphism
: S R such that and are the identity map.
In practice, a ring homomorphism : R S is an isomorphism iff is a
bijection.
Special case: R = S. A ring homomorphism : R R is called an endomorphism and an isomorphism : R R is called an automorphism.
Example.
Z
1. Z nZ
(where n is a positive integer) given by t t mod n is a hoZ
Z
momorphism: [0] is the zero in nZ
, [1] is the one in nZ
and if a, b Z,
then
(ab) mod n = (a mod n)(b mod n)

(this is the definition of the product in

Z
nZ ).

And similarly

(a + b) mod n = (a mod n) + (b mod n)

(this is the defintion of addition in

Z
nZ ).

So this is a ring homomorphism.

2. R = C with f : R R given by f (z) = f (z) is a ring homomorphism:

0 = 0, 1 = 1 and
f (a + b) = a + b = a + b = f (a) + f (b)
f (ab) = ab = ab = f (a)f (b).
Notice that since f is bijective, this is in fact an isomorphism (and an
automorphism). f is its own inverse, i.e. f f = identity.
3. R = R[x], S = R. Choose some R. Define : R[x] R by (f ) =
f (), where f (x) R[x] (e.g. if = 2 and f = x2 + 1, then (f ) = f (2) =
5). is called the evaluation homomorphism. Note that the polynomial 1
is not the same as the polynomial x: (x) = , but (1) = 1. is easily
checked to be a ring homomorphism.
4. The Frobenius Homomorphism: Say R is a commutative ring and say
Z
Z
or pZ[x]
). Define : R R by
p = 0 in R (e.g. R = pZ
(x) = xp = (x x x . . . x)
p times
(or (r) = rp ). This is also a ring homomorphism as (0) = 0, (1) = 1
and
(rs) = (rs)p = rp sp = (r)(s)


p
p
p
p1
(r + s) = (a + b) = a + pa b + +
api bi + + bp .
i




p
p
But p |
if 1 i p 1. Therefore
= 0 in R and hence
i
i
(a + b)p = ap + bp = (a) + (b).

5. R = Q( 2) = {a + b 2 : a, b Q}. Define : R R by

(a + b 2) = a b 2.

is
- (0) = 0

ring

homomorphism

because:

- (1) = 1

- ((a+b 2)(c+d 2)) = (ac+2bd+ 2(bc+ad)) = ac+2bd 2(bc+ad)

(a+b 2)(c+d 2) = (ab 2)(cd 2) = ac+2bd(bc+ad) 2 = (xy) = (x)(y).

6. Inclusions: Q , R, R , C, M2 (R) , M2 (C) are all ring homomorphisms.
Remark. Injective ring homomorphisms R S are the same as subrings of S.
Lemma 2.3. Let : R S be a ring homomorphism and let T be the image
of , i.e. T = {(r)|r R}. Then T is a subring of S.

10

Proof. By Lemma 1.5, we need to check that T contains 0,1 and that T is closed
under +, , . Clearly (0) = 0, (1) = 1, so 0, 1 T. Now say a, b T . Let
a = (r) and b = (s). Then
a + b = (r) + (s) = (r + s)
a b = (r) (s) = (r s)
ab = (r)(s) = (rs).
Hence T is a subring of S.

In fact, any map : X Y between sets factors as X Z Y with a

surjection and i an injection (Z Y is image of ). The above Lemma 2.3.
shows that the same is true for rings: any ring homomorphism : R S is

i
R T S, a surjection, i an injection and , i are ring homomorphisms.
We have already seen an example of a surjective ring homomorphism: : Z
Z/nZ, n 1.
Question: Are there any more surjections : Z R, where R is ring of a
completely different type to Z/nZ?
Answer: We will answer this soon.
Here is a problem that we need to solve first: Say : R S is a ring homomorphism. We have seen that Im() is a subring of S. Is it also true that ker() =
{r R : (r) = 0} is a subring of R (for example the kernel of Z Z/mZ is
the set of integers which are multiples of m = {. . . , m, 0, m, . . . })? No, as this
is not a subring of Z in general (for example it is very likely that 1
/ ker()).
Definition 2.4. A subset I R (where R is a ring) is called a left ideal if
(1) I is a subgroup of R (under +)
(2) If r R and i I, then ri I.
Similarily for right ideals. A subset I R (R a ring) is called a bi-ideal, or a
2-sided ideal, if I is a left and right ideal.
Remark. If R is a commutative ring, then all three of these notions coincide,
and we will call I an ideal, i.e. if
(1) I is a subgroup of R under +
(2) ri I for r R, i I.
Notation. If I is an ideal of R, we write I E R or I / R.
Example.
1. If R is a ring, then {0} and R are both bi-ideals of R.
2. Let R = R[x] be the set of all polynomials with real coefficients. Let I =
xR[x] be the polynomials with no constant term. If f = a1 x + +, g =
b1 x + + I, then f g = (a1 b1 )x + . . . has no constant term and
so I. Also, 0 I, therefore I is a subgroup of (R, +). Next, we need
to check that if f I and g R, then f g I (i.e. R is a commutative
ring). f = a1 x + a2 x2 + . . . , g = b0 + b1 x + . . . (b0 6= 0 is okay), then
f g = a1 b0 x + O(x2 ). Therefore, f g I, so I E R, i.e. I is an ideal of R.
11

3. Say m 1. Set I = mZ = {mt : t Z} R = Z (m is an integer), I is

the set of multiples of m.
Remark. I is the kernel of the map from Z Z/mZ.
Claim. I is an ideal. For 0 = 0m I. If a, b I, then a = ms, b = mt
for s, t Z. Therefore a b = m(s t) I, hence I is a subgroup of
(Z, +). Finally, if r R = Z, and i I, then i = mu for some u Z and
so ri = rmu = m(ru) I. Therefore, I is an ideal of R = Z (note that R
is a commutative ring).
Remark. If m = 0, then {mr : r Z} = {0} is also an ideal, and everything in the above proof works, giving another proof that {0} E Z.
4. All ideals of Z are of the form mZ = {mr : r Z} for m = 0, 1, 2, . . . .
For, say I Z is an ideal, then if I contains a negative number n < 0,
then (I is a subgroup) and n > 0 and n I. So either I = {0} or I
contains some positive integer. Let m be the smallest positive integer in
I. Easy check
I = {. . . , 2m, m, 0, m, 2m, . . . } = mZ.
For, certainly mZ I as m I and I is a group, and if t I, with t not
a multiple of m, then n such that
mn < t < m(n + 1)
and t mn I, but 0 < t mn < m, contradicting the definition of m
being the smallest positive integer in I. So I = mZ.
Ideals are hardly every subrings. In fact:
Lemma 2.5. If I is a left-ideal of R, and 1 I, then I = R.
Proof. Say r R. Then
I I = r1 I = r I,
so I = R.
Remark. This lemma (2.5) also holds for right ideals.
If R is a non-commutative ring, then we can define a new ring Ropp by Ropp = R
(as a set) and 0, 1 as before, with the same rule of addition, but a b in Ropp
is defined to be b a in R. So the left ideals of Ropp are the right ideals of R.
Definition 2.6. If : R S is a ring homomorphism, define the kernel of
to be
1 ({0}) = {r R : (r) = 0}.
Proposition 2.7. The kernel of a ring homomorphism is an ideal.
Proof. is by definition a group homomorphism (with the group law + and
identity 0), so the kernel of is a subgroup by M2PM2. Now say that r R
and i ker , i.e. (i) = 0. We want to show that ri and ir are in ker . But
(ri) = (r)(i) = (r) 0 = 0 = ri ker
by Proposition 1.8a. Similarly, ir ker . Hence ker is a bi-ideal.
12

Next, we will define the quotient ring R/I, where R is a ring and I is a bi-ideal
of R. This is well-defined, for if I E R is a bi-ideal, then I is a subgroup of
(R, +), so we can define the quotient group R/I per group theory. Recall that
an element of R/I is a subset r + I of R,
r + I = {r + i : r I}.
We will now aim to put a ring structure on R/I such that a natural map
R R/I is a ring homomorphism with kernel I.
Question: Is every bi-ideal the kernel of a homomorphism?
Set up: Let R be a ring and I E R a bi-ideal of R. Our goal is to form the
quotient ring R/I. So far we know that (R, +) is a group and I R is a normal
subgroup. Hence the quotient group R/I exists and has well-defined addition.
Recall that the elements of R/I are I-cosets in R, i.e. subsets of R of the form
r + I = {r + i : i I}.We will now make R/I a ring.
Define 0 of R/I to be 0 + I = I.
Define 1 of R/I to be 1 + I.
Define + on R/I to be
(r + I) + (s + I) = (r + s + I) .
By group theory, we know that this is well-defined.
For multiplication, define
(r + I)(s + I) = rs + I.
We need to check that this is well-defined. More precisely, that r0 = r + i, i I
and s0 = s + j, j I. Then r + I = r0 + I and s + I = s0 + I. So we need to
check that
rs + I = r0 s0 + I,
i.e. that is r0 s0 = rs + k, for some k I. Well,
r0 s0 = (r + i)(s + j) = rs + is + rj + ij.
Set k = is + rj + ij. We want to show that k I. Once we have checked that,
we are done. But i, j I and r, s R, so is, rj I since I is a bi-ideal. Also,
i, j I = ij I since I is also a left-ideal. So k K (as (I, +) is a group).
So we have a well-defined product on R/I induced from the product on R. We
now claim that R/I is a ring.
1. R/I is a group under addition, by group theory.
2.
(1 + I)(r + I) = 1r + I = r + I = (r + I)(1 + I),
so 1 + I works as the multiplicative identity. Moreover,
((r + I)(s + I)) (t+I) = (rs + I) (t+I) = rst+I = (r+I) ((s + I)(t + I)) .

13

Finally,
(x + I) ((y + I) + (z + I))

=
=
=
=

(x + I)(z + y + I) = x(y + z) + I
xy + xz + I
(xy + I)(xz + I)
(x + I)(y + I) + (x + I)(z + I)

and similarly for the other distributivity law.

Therefore, R/I is a ring.
Definition 2.8. Let R be a ring and I be a bi-ideal of R. We say that R/I is
the quotient ring.
Now, it is easy to check that the natural map R R/I given by r 7 r + I
is a ring homomorphism: the image of 0 is 0, the image of 1 is 1 and if r 7
r + I, s 7 s + I, then r + s 7 r + s + I and rs 7 rs + I. It is just as easy to
show that the kernel of R R/I is
= {r : r + I = I} = {r : r I} = I.
The First Isomorphism Theorem strengthens this. It says a surjective ring
homomorphsim is determined by its kernel.
Theorem 2.9. (First Isomorphism Theorem). Say that : R S is a homomorphism of rings. Say I = ker . This is an ideal of R by 2.7. Further let
T = Im. This is a subring of S by 2.3. Then there is a natural isomorphism
of rings
R/I
=T
(and indeed induces this natural isomorphism).
Proof. I is a bi-ideal and R/I is a well-defined ring. Our plan will be to define
a map : R/I T . Say r + I R/I. Define
(r + I) = (r).
Is this well-defined? Say r0 = r + i. Then r + I = r0 + I (this is if and only if).
Therefore we need to check (r) = (r0 ). But
(r0 ) = (r) + (i) = (r) + 0 = (r)
as I = ker . Hence is well-defined. Now, for injectivity of , say
(r + I) = (s + I) r, s R.
Then by definition of ,
(r) = (s) = (r s) = 0
since is a ring homomorphism. Therefore r s ker = I. So set r s = i.
Then
r = s + i = r + I = s + I.

14

by distributivity in R

Hence is injective. Next, surjectivity: say t T. We need to find r R such

that (r + I) = t. Well, T = Im(), and so r R such that (r) = t. Then
(r + I) = (r) = t,
so is surjective. Combining, is bijective. Now, set : T R/I to be the
inverse of . We leave it as an exercise to show that is a ring homomorphism.
Then + and are the identities. Therefore, is an isomorphism.
We saw already that the image of a ring is a ring. However, it is not true that
the image of an ideal is an ideal. For example, consider the map Z C given
by x 7 x. Then 2Z is an ideal of Z, but not of C.
However, the pre-image of an ideal is also an ideal. That is
Proposition 2.10. Say f : R S is a homomorphism of rings. Say I S is
a left ideal (resp. right ideal, resp bi-ideal). Then
f 1 (I) = {r R : f (r) I}
is a left ideal (resp. right ideal, resp. bi-ideal) of R.
Proof. Set J = f 1 (I). If , J, then f () I, f () I. Therefore, f (
) I. Hence J is closed under and 0 J and f (0) = 0. Therefore J is a
subgroup of R (under addition). Now say r R and j J. We need to show
that rj J (resp. jr J, resp rj.jr J). But
f (rj) = f (r)f (j) = f (r)f (j) I
S

(since I is a left ideal). Therefore f (rj) I = rj J (for the case of a right

ideal or a bi-ideal, the working is just the same).
Proposition 2.11. If R is a commutative ring, then R is a field R has
exactly two ideals (namely, {0} and R).
Say R is a commutative ring. What are all the ideals of R? We have seen that
if R = Z, then the ideals are {0} and nZ, n 6= 0. Another class of examples is
given by Proposition 2.11. above.
Remark. R = {0}, the zero ring, is not a field, by definition.
Proof. ( = ) Easy. Firstly note {0} =
6 R for a field, so there are at least 2 ideals.
Now say that R is a field and I R is an ideal and I 6= 0. We want to
show that I = R. Choose 0 6= x I. As R is a field, there exists y R
such that yx = 1. By definition of an ideal, 1 = xy I. Now say that
r R is arbitrary, then r = r1 I.
(=) We have two ideals, R 6= {0} (as otherwise, R = {0} has only one ideal).
We need to show that if 0 6= r R, then r has an inverse. So choose
0 6= r R. Set I = {ar : a R}. It is easily shown that I is an ideal.
Furthermore, r = 1r I, so I 6= 0, hence I = R, by assumption. Therefore
1 R = a such that ar = 1. Hence r has an inverse. So R is a field.

15

Therefore the only ideals of C are 0 and C (and similarly for R and Q).
Definition 2.12. Let R be a commutative ring. An ideal P of R is said to
be prime, or a prime ideal, if P 6= R and if a, b R with ab P, then either
a P or b P . An ideal M R is maximal if M 6= R and if J is an ideal with
M J R, then either J = R or M = J.
Proposition 2.13. R a commutative ring and I R an ideal. Then R/I is a
field if and only if I is maximal.
Proof.
(=) Say I is a maximal ideal. We want to show that R/I is a field. By
definition, I 6= R, therefore R/I 6= {0}. Now, need to show check that a
non-zero element of R/I has an inverse. So choose x + I R/I with x + I
not the zero element, i.e. x + I 6= I, i.e. x
/ I. We need to invert x + I in
R/I. Define a subset J R thus:
J = {ax + i : a R, i I}.
We claim that J is an ideal. We have that 0 J, since 0 R and 0 I.
Further, if ax + i1 and bx + i2 are in J, a, b R, i1 , i2 I, then
(ax + i1 ) (bx + i2 ) = (a b)x + (i1 + i2 ) J,
R

therefore J is a group under addition. Finally if r R, a R, i I, then

r (ax + i) = (ra)x + ri J.

Therefore J is an ideal.
Now, clearly I J (simply set a = 0). So I J R. But I is maximal.
Therefore J = I or J = R. But J 6= I, as x
/ I, but x J (a = 1, i = 0).
So J = R. Therefore 1 J, and so we can write 1 = ax + i for some
a R, i I.
We now claim that a + I is an inverse to x + I. For
(a + I)(x + I) = (ax + I) = 1 i + I = 1 + I = 1 of R/I.
( = ) Want R/I a field = I is maximal. Firstly, R/I a field = I 6= R.
Now say I J R and say J 6= I. We want J = R, then we are done. So
let us choose j J such that j
/ I. Then j + I 6= I in R/I (i.e. j + I 6= 0).
But R/I is a field. Hence j + I has an inverse, say k + I. Therefore
(j + I)(k + I) = 1 + I = jk I + i = jk + i = 1
for some i I. Finally, i I = i J and j J = jk J. Therefore
jk + i = 1 J. Hence J = R.
Proposition 2.14. R is a commutative ring, I is an ideal. Then I is prime iff
R/I is an integral domain.
Corollary 2.15. Maximal ideals are prime in a commutative ring.
16

2.13

obvious

Proof. M maximal = R/M is a field = R/M is an integral domain =

M is prime.

Proof. (of 2.14)

Case 1. I = R. Then I is not prime and R/I is not an integral domain. Now
say
Case 2. I 6= R. Then I is not prime iff a, b R such that ab I, a
/ I, b
/ I.
a, b R st. ab + I = 0 R/I
with a + I 6= 0, b + I 6= 0 in R/I. But
a + I, b + I R/I s.t. a + I 6= 0, b + I 6= 0 and (a + I)(b + I) = 0.
This is iff R/I is not an integral domain.
Remark. It is not hard to prove 2.15 directly (i.e. no quotients). We leave this
an an exercise.
Corollary 2.16. {0} is a prime ideal of a commutative ring R R is an
integral domain and {0} is a maximal ideal of R iff R is a field.
Corollary 2.17. Prime ideals are not always maximal.
Proof. There exists an integral domain that is not a field. for instance, Z, and
{0} is prime, but not maximal.
Corollary 2.18. Maximal ideals of Z are those of the form pZ, where p is
prime. Prime ideals of Z are of the form pZ, where p is prime, and {0}.
Proof. Z is an integral domain, but not a field. Hence {0} is prime, but not
maximal. All other ideals of Z are nZ, n > 0, where n is the smallest positive
1.11
element of the ideal after definition 2.4, and Z/nZ is a field it is an integral
1.16
domain n is prime.
Remark. We have seen that it is not true that prime = maximal. Is it true,
however, that prime and non-zero = maximal? This would be consistent with
everything we have seen so far. Yet, the answer is no. For example, consider the
set R = C[x, y] with I = {rx : r R}. Clearly I 6= 0 as x I. To check that I
is prime but not maximal, we need to check that R/I is an integral domain but
not a field. Well, consider the map C[x, y] C[y] given by f (x, y) 7 f (0, y).
It is easy to check that f is a surjective ring homomorphism with kernel being

the multiples of x. By the First Isomorphism Theorem, R

I = C[y], which is an
integral domain, but not a field (by Proposition 1.13).
Generators of Ideals

17

Definition. Let R be a commutative ring (out of sheer laziness) and X R a

subset. I want to talk about the ideal generated by X. We say that the ideal
generated by X in R is
\
I.
IR an ideal,XI
However, we will ignore this definition implicitly use
Lemma. If is a set and , I is an ideal of R, then
\
I=
I

is an ideal.
Proof. 0 I = 0 I. Also,
i, j I = i j I = i j I
i I, r R = i I = ri I = ri I.

Therefore there is a better definition: In the case that X is finite,

Definition 2.19. Let R be a commutative ring and X R be a finite subset
of R. Say X = {x1 , . . . , xn }. The ideal generated by X is the set
I = {r1 x1 + + rn xn : ri R}.
Notation. We write I = (x1 , . . . , xn ).
Lemma 2.20. I as defined above is an ideal and it is the smallest ideal of R
containing X.
Proof. 0 I (set ri = 0 i). Also I is closed under :
(r1 x1 + +rn xn )(s1 x1 + +sn xn ) = (r1 +s1 )x1 (r2 +s2 )x2 (rn +sn )xn .
Finally, if r1 x1 + + rn xn I and a R, then
a(r1 x1 + + rn xn ) = (ar1 )x1 + + (arn )xn I.
Furthermore, if J is any ideal of R with X J,
x1 , . . . , xn J = r1 x1 , . . . , rn xn J = r1 x1 + + rn xn J = I J.

Remark. We just showed that the ideal (x1 , . . . , xn ) is the smallest ideal of R
containing {x1 , . . . , xn }. Therefore
\
(x1 , . . . , xn ) =
I.
XIR

Hence both definitions (hard and easy one) are the same!
18

If X is infinite, the ideal generated by X is

{r1 x1 + + rn xn : xi X}
where the sum is finite, and n is as big as you like. Check that this is an ideal.
Special case: n = 1 and X1 = {x1 } = {x}, x R. Then (x) = {rx : r R} = Rx
is called a principle ideal.
Not every ideal is principal, for example, consider I C[x, y] defined by I =
(x, y), i.e.
I = {f x + gy : f, g C[x, y]}.
Check that I is the set of polynomials in x and y with no constant term. Further,
I is the kernel of the map C[x, y] C given by
f (x, y) f (0, 0).
We claim that I cannot be principal, as if I = (f ), then x I, then f divides
x and
= f = , 6= 0 or f = x, 6= 0
and y I = f 6= x. So f = 6= 0, but
/ I. Therefore (x, y) cannot be
principal.
However, if R = Z, and
I = (6, 8) = {6m + 8n : n, m Z},
then 2 I as 8 6 = 2, and therefore 2t I t Z, as I is an ideal. On the
other hand 6m + 8n is even for all n, m, and so I = 2Z = (2). Therefore, (6, 8)
is a principal ideal.
If fact, we have seen that every ideal of Z is of the form nZ = (n) for some
n Z. Hence every ideal is principal.
Question: Let a, b Z, not both 0. Let I = (a, b) be an ideal of Z. I must
be (d) for some d. What is d?
Definition 2.21. Let R be a commutative ring. We say that an ideal I is
finitely generated if
I = (x1 , . . . , xn ), xi R.
We say that I is principal if I = (x) for some x R. Further, we say that R is
Noetherian if all ideals of R are finitely generated. Finally, we say that R is a
principal ideal domain (PID) if
1. R is an integral domain
2. all ideals of R are principal.
We think of these definitions in the following way:

19

- R noetherian R is finite dimensional. Therefore R not Noetherian

= R pathological (much too big).
- R a PID = R is one-dimensional
Here are some things we cannot prove yet:
- C[x] is a PID
- C[x1 , . . . , xn ] is not a PID if n > 1. But it is Noetherian
- C[x1 , . . . ] is not Noetherian.
We have proved that Z is a PID, as it is clearly an ID and all ideals are of the
form nZ, and hence principal.

20

Chapter 3

Factorisation in Integral
Domains
The purpose of this chapter is to axiomatise and generalise the proof that any
n Z 2 is uniquely a product
of primes. Itwill turn out that an analogous
theorem is true in Z[i] and Z[ 2], but not Z[ 5].
Throughout this chapter, R will denote an integral domain (so R is commutative).
Definition 3.1. Say x R is a unit if y R such that xy = 1. Write R =
the set of units in R (note that R is a group under multiplication with identity
1). We say x divides y (denoted x | y) if q R such that y = qx. We call x
and y associatives if y = ux for some unit u R.
Exercise. Show that x R is a unit (x) = R. Show also that x | y
y (x) (y) (x).
Note. The notion of divides is the usual one in, for example, Z or R[x] etc.
Lemma 3.2. x and y are associatives (x) = (y).
Proof. First note that if y = ux and u R , then v such that uv = 1 and
vy = vux = x. Note that the notion of being associatives is symmetric. If x = 0,
then ux = 0, and the only associative of 0 is 0 = 0 1. On the other hand, if
(0) = (y), then y (0) = {0}, and so y = 0. Hence the lemma is true for x = 0
(or y = 0, by symmetry). Now say x 6= 0. Then y = ux, u a unit. = x = vy
(where v = u1 ) and then (x) = Rx = Rvy = Ry = (y) as Rv = R. Conversely,
if (x) = (y), then x (y) = x = ry and y (x) = y = sx. Hence
x = rsx = x(rs 1) = 0
and as x 6= 0 and R is an integral domain, = rs 1 = 0 = r R and x
and y are associatives.
Corollary. Being associatives is an equivalence relation.

21

Example. R = Z. The units of Z are {r Z : r | 1} = {1}. Hence the

associatives of n are n.
Definition 3.3. We say r R is irreducible if r 6= 0, r not a unit, and if r = ab,
then either a or b is a unit (as an example, R = Z : irreducible = usual notion
of prime, up to sign: r Z irreducible r = p, p prime). We say r R
is prime if r 6= 0, r 6= an unit, and if r | ab = r | a or r | b.
Exercise. Assume that the integers factor uniquely into primes. Check that
the primes of Z are exactly p, for p a prime number.
Example. _
1. Let R = Z. 2 is irreducible (as is any prime number) and 3 is too.
2. R = Z[i] = {a + bi : a.b Z}. The units of R are found as follows:
2
Define N : R Z0 by N (a + ib) = a2 + b2 , i.e. N (z) = zz = |z| .

It is easy to see that N (rs) = N (r)N (s). Say r R , i.e. s such

that rs = 1. Then N (r)N (s) = 1 = N (r) = 1 as N (r) Z0 .
So if r = a + ib and r R , then a2 + b2 = 1 = r = 1, i.
Conversely, 1 and i are units. But 2 Z[i] is no longer irreducible,
because 2 = (1 + i)(1 i), which is a product of two non-units (hence
irreducibility of 2 depends on R). But 3 is still irreducible in Z[i], as if
3 = rs, r, s Z[i], then 9 = N (3) = N (r)N (s). Let r = a + bi, s = c + di.
Then (a2 + b2 )(c2 + d2 ) = 9 = a2 + b2 {1, 3, 9}. But a2 + b2 = 3 has
no solutions in Z. So either N (r) = 1 or N (s) = 1. So r or s is a unit.

3. Consider R= Z[ 5] = {a + b 5 : a, b Z}. Define N : R Z0

by N (a + b 5) = a2 + 5b2 (= zz). If r is a unit, then rs = 1for some
s R. Then N (rs) = N (r)N (s) = 1 = N (r) = 1. r = a + b 5 and
a2 + 5b2 = 1. So b = 0, a = 1. Hence r = 1 and these are both units.
Here, there is no solution to a2+ 5b2 = 2 or a2 + 5b2 = 3 with a, b Z.
Hence 2, 3 are irreducible in Z[ 5]. For example,
2 = rs = 4 = N (2) = N (r)N (s) = N (r) {1, 2, 4} N (r) = 1 or N (s) = 1

and 1 + 5 is irreducible as N (1 + 5)= 6 and factors of 6 in Z1 are

1,2,3,6 and 2,3 are not possible.
Also

1 5 is irreducible (norm is also

6). Now 6 = 2 3 = (1 + 5)(1 5), i.e. there are two factorisations
of 6 into irreducibles. On the
other hand,
6 has no factorisations into
primes, because none of 2,3,1 + 5, 1 5 are prime. For example, 2
divides

6 = (1 + 5)(1 5)

1 5
2

but2 - 1 + 5, 1 5 (as

/ R) etc. Hence we see that Z[ 5]

has irreducibles that are not prime and some elements factor into irreducibles in more that one way.
On the other hand,
Lemma 3.4. All primes are irreducible in an integral domain.

22

Proof. Say r is prime. Then r 6= 0 and r is not a unit. Say r = ab. We will
show that one of a, b must be a unit. Now, r = ab = r | ab. But r is prime,
hence, wlog, r | a (could be r | b as well). So
a = sr, s R = r = ab = srb = r(bs 1) = 0.
As r 6= 0, we must have that bs = 1, since we are in an integral domain, and
hence b is a unit.
Lemma 3.5. If 0 6= r R, then r is prime (r) is a prime ideal.
Proof. If r is a unit, then r is not prime and (r) = R is not a prime ideal. So
say that r is not a unit. Then (r) is a prime ideal ab (r) = a
(r) or b (r) r | ab = r | a or r | b. This is equivalent with saying that
r is prime.
Definition 3.6. An integral domain is called an Euclidean domain (ED) if there
is some function : R\{0} Z0 such that
(1) (ab) (a) if a, b 6= 0
(2) If a, b R and b 6= 0, then one can write a = qb + r with q, r R (we call q
the quotient and r the remainder) such that either r = 0 or (r) < (b).
Example.
1. R = Z and (r) = |r| .
2. For F a field, R = F [x], (r) = deg(f ).
Exercise. Verify whether the last two examples are indeed Euclidean domains.
Theorem 3.7. R is an Euclidean Domain = R is a prime integral domain.
Remark. Compare this with the proof that all ideals of Z are principal.
Proof. Say that R is an Euclidean domain and I R is an ideal. If I = {0} =
(0) than this is clear. Now, assume that I 6= {0}. Choose n I\{0} with (n)
minimal. We claim that I = (n). Certainly (n) I. Now say i I. We want to
show that i (n). For (i = a, n = b), we can write i = qn + r with either r = 0
or (r) < (n). But r = i qn I. Hence (r) < (n) cannot be true by the
definition of n. Hence r = 0 and so i = qn (n). Hence I = (n).
Our next goal is to show that things factor uniquely into primes if we are in an
prime integral domain.
Corollary 3.8. F a field = F [x] is a PID.
Proof. Obvious.

At this stage, recall that in Z, prime = irreducible, but in Z[ 5], prime 6=

irreducible. In an general integral domain R, prime = irreducible.
Lemma 3.9. In a PID, all irreducibles are prime.
23

Proof. Say R is a PID and r is irreducible. Then r 6= 0 and r is not a unit. Say
r | ab, a, b R and assume r - a. We want to show that r | b. Define I = (r, a).
As R is a PID, we must have that I = (x) for some x R. So r, a (x), and so
r = sx and a = tx. But r is irreducible, therefore either s or x is a unit. But s
cannot be a unit. For if s is a unit, su = 1 for u R and
r = sx = ur = x = a = bx = tur = r | a
contradiction. So x must be a unit. Hence I = (x) = R = i I, and therefore
, R such that r + a = 1 = b = rb + ab and r | rb, r | ab (as
r | ab). Hence r | b.
Definition 3.10. An integral domain R is a unique factorisation domain (UFD)
if
UF1 (factorisation) Any non-zero r R can be written r = ur1 . . . rn for some
n 0 with u a unit and ri irreducible
UF2 (uniqueness) If r = ur1 . . . rn = vs1 . . . sm with m, n 0 with and u, v
units and ri , si irreducibles, then m = n and after reordering the si , if
necessary, ri and si are associates i.
Remark. UF2 is necessary to deal with, e.g, 15 = 3 5 = 5 3 = 3 5 =
1 3 5 etc.
Example.
1. Z is a UFD.
2. F [x] is a UFD.
3. Any ED is a UFD.
Remark. We have seen that in any PID, prime = irreducible. This is, more
generally, true in a UFD:
For prime = irreducible in an ID (shown before). For the converse, say r
is irreducible. Then r 6= 0 and r is not a unit. Hence we only need to check
r | ab = r | a or r | b. So say r | ab. If a = 0 = r | 0 = done. Say
a, b | 0.
Say rs = ab. Factor s, a, b :
- s = us1 . . . sm
- a = va1 . . . an
- b = wb1 . . . bp
where u, v, w are units and si , rj , bk irreducible. Now, get two factorisations of
rs = ab :
rs = us1 . . . sm r = vwa1 . . . an b1 . . . bp .
By UF2, these two factorisations are the same up to order and associates. Hence
r is an associate of some ai or some bj . Wlog, say ai = ur. Then r | ai | a =
r | a. Hence prime = irreducible in a UFD.
24

Remark. We have seen that ED = PID and we will see that PID = UFD.
The converses, however, are both false. In fact it is a theorem that if R is a
UFD, then so is R[x]. In particular, we see that C[x, y] is a UFD: C is a field
C is a PID (the only ideals are (0) and (1)) Hence C[x] is a UFD, and so C[x, y]
is a UFD.
But the ideal (x, y) is not principal. In fact, we have that PID = dim 1,
and C[x, y] has dim 2.
It is much harder to find a PID that is not an ED.
i
h

Example. Z = 1+ 219 = {a + b : a, b Z}. Note,

2

1+

19

2

18 + 2 19
9 + 19
=
=
= 5.
4
2

= Z[] is a ring. By M3P15, this is also a PID, and, by a messy calculation,

is it not a ED.
Example. Z[i] is a UFD. It suffices to prove that Z[i] is an ED. Define (a +
ib) = a2 + b2 . We need to check that if x, y Z[i], then x = qy + r with
(r) < (y) or r = 0.
How to find q : consider

x
y

C. So Z[i] is an ED Z[i] is a UFD.

Exercise. _

1. Show that Z[ 2] is a UFD.

2. Why does the above procedure fail to work for Z[ 5]?

Theorem 3.11. A PID is a UFD.
Proof. Say R is a PID. Assume UF1 fails. Choose r R such that r 6= 0
and r 6= ur1 . . . rn , ri irreducible for i = 1 n. ||||| Clearly r is not a unit
(n = 0) and r is irreducible (n = 1). Hence r = r1 s1 for some r1 , s1 , not units
in R. If r1 = ut1 . . . tn and s1 = vw1 . . . wm , ti , wj irreducible and u, v units,
then r = uvt1 t2 . . . w1 . . . wm is a factorsation of r. This is a contradiction to
the definition of r. Hence one of r1 , s1 does not factor into irreducibles either.
Wlog, take r1 . By the same trick, r1 = r2 s2 where s2 is not a unit and r2 is
not the product of irreducibles. Similarly, r2 = r3 s3 , s3 not a unit and r3 not
the product of irreducibles. By repeating this procedure, we obtain an infinite
sequence
r = r0 r1 . . .
where ri = ri+1 si+1 , si+1 not a unit. Now, r = r1 s1 ideal (r1 ) contains r and
hence (r). Furthermore, (r) (r1 ) (for if (r1 ) = (r), then
r1 = rt, t R, = r = r1 s1 = rts1 = r(1ts1 ) = 0 and r 6= 0 = s1 is a unit
contradiction). Similarly, (r1 ) (r2 ) with (r1 ) 6= (r2 ), and so we get an increasing chain of ideals
(r0 ) (r1 ) . . .
25

where all containments are strict. Now, let I =

i, j I, then N  0 such that

n0 (rn ).

I is an ideal, since if

i, j (rN ) = i j (rN ) I.
As R is a PID, I is pricipal. Hence I = (d) for some d R and d I = d
(rN ) for some N 0. Therefore
(d) (rN ) ( (rN +1 ) ( I = (d)
contradiction. Hence PID = UF1.
For PID = UF2, consider lemma 3.9: irreducibles are prime in a PID. So,
as r 6= 0 and r = ur1 r2 . . . rn = vs1 . . . sm . We will prove that n = m and after
re-ordering ri and si are accociates by induction on n. If n = 0, then r = u is
a unit and if m > 0, then s1 | u = s1 = 1. But s1 is irreducible, hence s1
is not a unit, contradiction. Hence m = 0, and so the base case of induction
holds. Now the inductive step: Let n 1 and assume the statement is true for
n0 < n. Then r = ur1 . . . rn , n 1 and r = vs1 . . . sm . So r1 | r = vs1 s2 . . . sm
and r1 is irreducible, hence r1 is prime by 3.9. Therefore r1 | v or r1 | si for
some i. As v is a unit v | 1, so r1 | v = r1 | 1, contradiction. Hence r1 | si
for some i, 1 i m (and in particular m 1). After re-ordering the, wlog,
si , r1 | s1 . Say s1 = r1 t for some t. s1 is irreducible, hence either r1 or t must
be a unit. But r1 is not a unit (see above), so t must be a unit. Hence r1 and
s1 are associates. Now, cancel r1 (which is fine, as R is an ID). So
ur1 r2 . . . rn = vs1 s2 . . . sm = vr1 ts2 . . . sm

= ur2 . . . rn = vt s2 . . . sm
unit
and by our inductive hypothesis, we must have n 1 = m 1 = n = m and
ri and si are associatives for all i 2 after re-ordering, if necessary.
Remark. As a consequence, if n 1 and t Z such that t2 1 mod n, then
n = a2 + b2 , a, b Z.
Remark. If p is prime and p 1 mod 4, then t such that t2 1 mod p,
for example t = p1
2 . To show this we could, alternatively, use the fact that

(Z/pZ) is cyclic elements of order 4 (namely t).

h i
 

Note. Z[ 3] = Z 1+ 2 3 = algebraic integers in Q( 3) = a+b 1+ 2 3 , a, b

Z.
Questions:

(Q1) What are the algebraic integers in Q( d), d Z not being a multiple of
a square number?

(A1) This
will be answered in M3P15: Z[ d] if d 6 1 mod 4 or
h question
i
Z 1+2 d if d 6 1 mod 4.
26


(Q2) When is the set of algebraic integers of Q( d) an ED, PID, UFD?
More generally,
what about algebraic integers in a general number field

like Q( 3 2) etc.?
(A2) This question is fully answered for d < 0 : There are only 9 values of
d < 0 which give rise to rings which are UFDs (i.e. generally, this is not
an UFD (and therefore, not a PID or ED). The largest value of |d|
which
givesh rise to aiUFD is d = 163 and the algebraic integers in Q( 163)
is Z

1+ 163
2

. 1+

163
2

is a root of

x2 + x + 41 = 0
which is prime for 0 x 39. It was known ages agothat there exists a d
with d < 163 such that the algebraic integers of Q( d) is a UFD, but it
was not proved until the 1960s thatno further such ds exist. For d > 0,
it was conjectured by Gauss that Z[ d] was a UFD infinitely often. This
conjecture is still open.

(Q3) Another question is whether the hintegers

of Q( d) form a UFD infinitely
i

often. The natural norm on Z d , d > 0, given by (a + b d) =


 2
a db2  is a good canditate for the Euclidean norm.
(A3) It can be shown that this is not anEuclidean norm for d bein sufficiently
large. An interesting example is Z[ 14]. This is a UFD and a PID, but is
it an ED? It was known for
a while that the generalised Riemann hypothit was
esis would imply that Z[
 proved
14] is an ED. However, in 2004

unconditionally that Z[ 14] is an ED (note that (a+b 14) = a2 14b2 

does not yield an ED).
Remark. Some final facts about UFDs: We have seen an example
of a ring that
5]): 2 3 =
failsto be a UFD,
because
uniqueness
of
factorisation
fails
(Z[

(1+ 5)(1 5). But clearly (as it can

be
shown
using
induction
on a2 +5b2 )

every non-zero, non-unit element of Z[ 5] is a product of irreducibles: only

uniqueness happens to fail.
We will end this chapter with an example of a ring which has elements which
do not even factorise into irreducibles (i.e, UF1 fails rather than UF2).
Definition 3.12. Let R be an integral domain. We say that a properly divides
b if
(1) a | b
(2) a, b are not associates (i.e. b = sa, s not a unit).
Example. 3 divides 3 in Z, so 3 properly divides 6 in Z.
Lemma 3.13. In a UFD R, you cannot have a sequence r1 , r2 , . . . of non-zero
elements of R such that ri+1 properly divides ri i (c.f. proof of UF1 for a
PID).

27

Proof. Clearly r1 6= 0 = r1 = u1 . . . n where i is irreducible and u is a

unit. Hence r2 has at most n 1 irreducible factors, and continuing, we find
that rn must be irreducible and rn+1 must be a unit. But then there cannot be
an rn+2 properly dividing rn+1 .
Example. Now let R be the subring of Q[x] consisting of all polynomials f =
a0 + a1 x + + an xn with a0 Z. Note that R is an integral domain (as
/ R). Now look at the sequence
R Q[x], which is an ID) and 2
/ R (as 12
x x
x, 2 , 4 , . . . . It is easy to see that each element properly divides the previous one.
Hence R cannot be a UFD. Indeed, we cannot factor x R into irreducibles:
x
x=NN
for all N 1. This is a problem. In fact, reversing the implication
PID = UF1, we can prove that R is not Noetherian. Indeed, the ideal
I R of polynomials with zero constant term is not finitely generated: If
f1 , f2 , . . . , fn I and J = (f1 , . . . , fn ) I is the ideal generated by all the fi
and if fi = ai x + . . . , then ai Q, and so N  0 in Z such that N ai Z i.
1
x I, but
Check that f J, f = ax + O(x2 ) and N a Z. Hence J 6= I, as 2N
not in J.

28

Chapter 4

Localisation
Let us start with a special case: the field of fractions of an Integral Domain.
We have seen the construction of R/I, where R is a ring and I a bi-ideal. Also,
R/I is smaller than R. The only way to make bigger rings, that we have seen
so far, is by adding a polynomial variable R R[x] R. In this chapter we
will see another way of creating bigger fields out of smaller ones, e.g. Z , Q.
As an example, consider a transcendental number in C, i.e. some x C such
that x is no non-zero root of any polynomial in Q[x], e.g. x = . Consider the
subring Q[] = {a0 + a1 + + an n : ai Q} of C. As abtract rings, we have
Q[]
= Q[x]. But this is not a field. We want
T to ask the question: What is the
smallest subfield of C that contains (i.e. F C,F,F field F )? It is clear that
it most contain 0 and 1 and so it must contain Z, and as it is a field, it must
also contain all elements of Z1 etc. As it contains Q and , it must also contain
1
Q[]. Since 1 and in fact p()
Q[] for any p Q[x], p 6= 0 ( = p() 6= 0).
Hence it must also contain
{z C : z =

a()
b()

a()
b()

for any a, b Q[x], b 6= 0. This works through:

for a, b Q[x], b 6= 0} is indeed a field. This worked because

Q C and C and C is a field. Hence

a()
b()

makes sense in C.

But what happens if you do not have a field?

We will try to answer the following question: If R is an integral domain, how do
we build a field F such that R F and F is (somehow) generated by R? For
example, Z Q and Q[] Q() = { a()
b() : a, b Q[x], b 6= 0} are all within C.
the construction is called the field of fractions.
Theorem 4.1. (Field of fractions of an integral domain). Let R be an integral
domain. Then a field F and an injective ring homomorphism : R F such
that for any f F, a, b R, b 6= 0, such that f = (a)
(b) . Furthermore, the pair
(F, ) is unique up to unique isomorphism, i.e. if (F, ) and (F 0 0 ) both work,
then there is a unique isomorphism of fields i : F F 0 such that

29

>F
}}
}
}}
}}

i
RA
AA 0
AA
AA

F0

and the maps commute, i.e. i = 0 .

Proof. Let X = R (R\{0}) (think of this cross-product as a pair of numerator
and denominator). Define a relation on X by (a, b) (c, d) ad = bc.
Claim. is an equivalence relation.
1. Reflexive: Obvious
2. Symmetric: Obvious (as R is an integral domain, hence commutative)
3. Transitive: Say (a, b) (c, d) (e, f ). We want to show that (a, b)
(e, f ). Now ad = bc and cf = de, and so adf = bcf = bde and d 6= 0 by
assumption. Hence af = be by cancellation in an integral domain, and so
(a, b) (e, f ).
Define F to be the set of equivalence classes and denote the equivalence class of
(a, b) by [a : b]. Next, we make F a ring:
Define 0 = [0 : 1] = [0 : t]t 6= 0 and 1 = [1 : 1] = [t : t] t 6= 0. Further, let
[a : b] + [c : d] = [ad + bc : bd]. Is this well-defined? Say [a : b}] = [A : B] (i.e.
(a, b) (A, B), is [a : b] + [c : d] = [A : B] + [c : d]?). Well, need to check
aB = bA = [ad + bc : bd] = [Ad + Bc : Bd],
i.e. that (ad + bc)Bd = bd(Ad + Bc). Therefore, we need to check
aB = bA = aBd2 + bcBd = Abd2 + Bcbd
and this is clear. Define [a : b] [c : d] = [ac : bd]. This is also well-defined
(exercise!). Now, we only need to check the axioms:

30

Additive Inverse: Note that [a : b] + [a : b] = [ab ba : b2 ] = [0 : b2 ] = 0

Associativity:
([a : b] + [c : d]) + [e : f ] = [ad + bc : bd] + [e : f ] = [adf + bcf + bde : bdf ]
and
[a : b] + ([c : d] + [e : f ]) = [a : b] + [cf + de : df ] = adf + bcf + bde + bdf ]
+ is associative (and clearly also commutative). Hence (F, 0, +) is a
group. Also is associative and 1 is the identity.
Distributivity We need [a : b]([c : d] + [e : f ]) = [a : b][c : d] + [a : b] [e : f ].
But the left hand side is
[a : b][cf + de : df ] = [acf + ade : bdf ]
and the right hand side is
[ac : bd] + [ae : bf ] = [acbf + bdea : b2 df ]
and we can cancel b because it is a denominator.
We will show that these equivalence classes form a ring F (in fact, F is even a
field): [a, b] = [0 : 1] a = 0). Therefore, if [a : b] 6= [0, 1] then a 6= 0 and
inverse is [b : a]. It is easy to see that [a : b][b : a] = [1 : 1]. : R F defined
by (a) = [a : 1] is a ring homomorphism (easily checked). Also is injective:
the kernel is an ideal and (a) = 0 = [a : 1] = [0 : 1] = a = 0. So it is
injective.
So we see that if R is an integral domain, then F is a field and : R , R. Now,
given f F, f = [a : b], b 6= 0, we have f = (a)
(b) because
(a)
= (a)(b)1 = [a : 1][b : 1]1 = [a : 1][1 : b] = [a : b].
(b)
All that is left is to show that F is unique in the following strong sense: if F 0
is any other field, and R , F is an injective ring homomorphism such that
0
f 0 F 0 , f 0 = (a)
(b) , for some a, b 6= 0, then a unique : F F such that
>F
}}
}
}}
}}

RA
AA 0
AA
AA

F0

commutes.
For uniqueness of , say f F. Then f =
know ((a)) = 0 (a), ((b)) = 0 (b) and

(a)
(b)

for some a, b 6= 0 a, b R. We

(f ) 0 (b) = (f )((b)) = (f (b)) = ((a)) = 0 (a).

31

 0 (a)
(a)
0 (b) for f = (b) . Finally, we claim
0 (a)
write f = (a)
(b) , and set (f ) = 0 (b) . Need to
(c)
0 (a)
0 (c)
if f = (a)
(b) = (d) , b, d 6= 0, then 0 (b) = 0 (d) .

So if exists it must be defined by (f ) =

that we can define by if f F,
check that this is well-defined, i.e.
Well,

(a)
(c)
=
= (a)(d) = (c)(b) = (ad) = (cb) = ad = bc
(b)
(d)
as is injective. So 0 (a)0 (d) = 0 (b)0 (c) =
we have that (b), (d) 6= 0).

0 (a)
0 (b)

0 (c)
0 (d)

(as 0 is injective,

Now is a ring homomorphism (easy to check). Is an injection? Yes, F is

a field = ker() is an ideal, but the only ideals of a field are 0 and F . But
(1) = 1 (easy) = ker() 6= F = ker() = 0.
Last thing to check: Is a surjection? Yes, because if f 0 F 0 , then by assump0
0
tion f 0 = 0(a)
(b) for some a, b, b 6= 0. Therefore f = ([a : b]).
Notation. F is called the field of fractions of R and it is denoted by F rac(R).
Corollary 4.2. Any integral domain is a subring of a field.
One can do a slightly more general construction here and go from Z to a smaller
ring than Q, i.e. Z a ring Q (there are lots of rings that meet this
requirement).
Definition 4.3. Let R be a commutative ring. A multiplicative subset of R is
S R if
(1) 1 S
(2) s, t S = st S.
Example. _
1. Let R be a commutative ring, a R. S = {1, a., a2 , . . . }.
2. Let R be a commutative ring, P R a prime ideal. S = R\P (also, could
have S = {I}, I any ideal).
3. A special case of 2.: R is an integral domain, P = (0), S = R\{0}.
Definition 4.4. Let R be an integral domain and let S be a multiplicative
subset and assume 0
/ S (e.g. S = R\{0}) . We define the localisation of R at
S to be the equivalence classes [r : s] R S such that
(r1 , s1 ) (r2 , s2 ) r1 s2 = r2 s1 .
In other words, the localisation of R at S is either
(1) the equivalence classes of R S under (r1 , s1 ) (r2 , s2 ) r1 s2 = r2 s1
where +, 0, 1 are defined as in F rac(R) (special case: S = R = {0})
32

(2) subset of F rac(R) consisting of elements of the form r/s for r R, s S.

Notation. We denote the localisation of R at S by RS or S 1 R.
Example.
1. S = R\{0}, we get F rac(R).
2. R = Z, p prime = S = {1, p, p2 , . . . }. Then the localisation is { pan :
a Z, n 0} Q.
3. R = Z, P = (p), S = Z\P = {n Z : p|n}. Then the localisation is
{ ab : p - b}.
Note. There is an injection R RS given by r r/1 (that is why we demanded
1 S).
Lemma 4.5. Let R an integral domain and 0
/ S a multiplicative set. If A is
any commutative ring and if : R A is any ring homomorphism such that
(s) is a unit in A s S, then 0 : RS A extending (i.e. such that
R

RS

/A
>

and 0 is unique.
Proof. Uniqueness of 0 : if r/s = [r : s] RS then 0 (r/s) 0 (s) must equal
0
(r)
(r)
0 (r). This forces 0 (r/s) =
0 (s) = (s) . Conversely, this does work. Also
0 (r/s) =
r
s

r0
s0 ,

(r)
(s)
0

makes sense in A, as (s) is a unit by assumption. Now, if

then s r = r0 s and so
(s0 )(r) = (r0 )(s) =

(r0 )
(r)
=
.
(s)
(s0 )

Hence 0 is well-defined. It is easy to show that 0 is a ring homomorphism.

Remark. If R is an integral domain and P R is a prime ideal and S = R\P ,
then people often write RP for RS , S 1 R.
Example.
(1) S = R\{0} (null subset as R is an integral domain), then S 1 R = F rac(R)
(2) S = {1}, then S 1 R = R
(3) p prime and R = Z, S = {1, p, p2 , . . . } (write this as S 1 Z = Z[ p1 ]) then S 1 R =
{ pan : a Z, n 0} which is a ring, a subring of Q.
Exercise. If q is a prime number, q 6= p, then (q) S 1 R is still a prime
Z
ideal as Z qZ
given by p something non-zero. But Z/qZ is a field. Hence
p a unit. So Z Z/qZ extends to a map S 1 Z Z/qZ and one can
check that the kernel is (q). Hence (q) is maximal ((q) = qS 1 Z = { pan : a
Z multiple of q, n 0}). But p is a unit in S 1 Z. Hence (p) is no longer prime
(Spec(S 1 Z) Spec(Z) is an open set. The complement is (p)).
33

(4) R = Z, P = (p), p prime, P a prime ideal generated by p. S = Z/P = {n Z :

p - n}. Now S 1 R = { ab : a Z, b Z and p - b}. S 1 R is a ring called Z(p) , Z
localised at p. If q is prime, q 6= p then q S 1q Z(p) = q is a unit in
Z(p). But p1
/ Z(p) and indeed pZ(p) is a prime ideal and a maximal ideal: a
map Z(p) Z/pZ extending Z Z/pZ as any s S is a unit mod p, and the
kernel of this is (p).
In fact, Z(p) has exactly two prime ideals, namely (0) and pZ(p) , one contained
in the other.
Remark. Spec(Z(p) ) = 2 points, one closed and one open.
Exercise. Show that Q has uncountably many subrings.
Definition 4.6. A commutative ring is local if it has a unique maximal ideal.
Example. A field has only two ideals, 0 and itself. But by definition a field
cannot be maximal in itself, hence 0 is the maximal ideal.
Let us construct some more interesting local rings:
Zorns Lemma
This is an axiom of mathematics (it is completely equivalent to the axiom of
choice). A poset or partially ordered set is a pair (S, ) where S is a set and
is a binary relation such that
(1) s S, s s
(2) if s t and t s, then s = t
(3) if s t and t u, then s u.
Remark. partial because we do not assume s, t either s t or t s.
Example. A good example to bear in mind: if S is a subset of X, then s
t s t.
A chain in a poset is a subset T S if t1 , t2 T either t1 t2 or t2 t1 .
Example.
- The empty set is a chain.
- {s} is a chain
- If S = set of subsets of X and s1 s2 s3 . . . , then {sn } is a chain.
Definition. An upper bound for a subset W S is s S such that w W
we have w s. A maximal element of a poset S is x S such that if y S and
x y, then x = y.
Remark. I did not just say that x is maximal if z, z x. It is fine to have
x S maximal and z S such that z x and x z (maximal 6= bigger than
everything but maximal = nothing bigger).

34

Example. s1 s2 . . . subsets of X, then s =

and X is too.

n1 sn

is an upper bound

Lemma 4.7. (Zorns Lemma) Say S is a poset with the property that every
chain has an upper bound. Then S has a maximal element (possibly many
maximal elements).
Here is why we want Zorn:
Theorem 4.8. If R is a ring and I R is an ideal with I 6= R, then maximal
ideal m such that I m R.
Proof. S = {ideals J R such that I J, J 6= R}. Define J1 J2 J1
J2 . Note that I S S 6= 0 (which is good, as the empty chain has an upper
bound, namely I). If T S is a non-empty
S chain (T is a big collection of ideals
totally ordered by inclusion), then J = xT x. J is an upper bound for T as
x J x T and I J and furthermore J 6= R, as if J = R, then 1 J. Then
1 x for some x T = 1 x for some x S. But 1 x = x = R and
R
/ S. Hence chains have an upper bound by Zorns Lemma, S has maximal
elements and it is easy to check that these are maximal ideals.
Corollary 4.9. If R 6= {0}, then R has a maximal ideal.
Proof. I = {0} and then use 4.8.
Proposition 4.10. Say R is a commutative ring and I R is an ideal. Then
I is the unique maximal ideal of R iff R is the disjoint union of I and R .
Proof. Say I is the unique maximal ideal of R. Then I 6= R, so I = is
not a unit (as a unit = 1 I = I = R). Conversely, if
/ I, then let
J = (). We want J = R, as then 1 J = R such that = 1 and
is a unit. But if J 6= R, then by 4.8, J m, the maximal ideal and so I is the
unique maximal ideal. Hence m = I = J I = J I, contradiction.
Conversely, say R is the disjoint union of R and I. Say J is any ideal, J 6= R.
Then J = is not a unit = I. Hence J I and so I is the
unique maximal ideal.
An Example of a Local Ring
Let R be an integral domain, P R be a prime ideal. Let S = R\P.
Claim. S 1 R is a local ring and the unique maximal ideal is { as : a P, s
S} = S 1 P.
Proof. First check that S 1 R is an ideal:
a b
at + bs
+ =
,
s
t
st

st S, a, b P = at + bs P

and as S 1 P , 0 = 01 S 1 P . Also as S 1 P and bt S 1 R = ab

st
S 1 P as ab P, st S. Finally, if S 1 R and
/ S 1 P, then I claim that
/ P = a S = as S 1 R.
is a unit. For = as , s S, R and
a
Hence s is a unit.
35

Example.
1. Z(p) is a local ring with unique maximal ideal { ab : p | a, p - b}, the quotient
field Z\pZ.
2. R = C[x, y] (polynomials in two variables). P = (x, y) = kernel of
evaluation map R C, F F (0, 0). Rp = { fg : f, g C[x, y], g
/
(x, y), i.e. g(0, 0) 6= 0}. Therefore RP = rational fuctions (ratio of two
polynomials) on C2 which are well-defined in a neighbourhood of (0,0).
(0,0)
.
The evaluation map RP C is given by fg fg(0,0)

36

Chapter 5

More on Polynomial Rings

Throughout this chapter, R will be a commutative ring. Recall that R[x] =
{a0 + + an xn : ai R, n 0 arbitrary}. The purpose of this chapter is to
determine which, good, properties of R are inherited by R[x]. For example, if
R is a field, then this does not imply that R[x] is a field (as x has no inverse).
But R[x] is, at least, a PID (as it is a ED). Also, if R is an ID, then R[x] will
also be an ID (as proved in 1.13). We will now prove something deeper: we
will show that if R is noetherian, then R[x] is also noetherian. Recall that R is
noetherian if all its ideals are finitely generated, for example a field or a PID.
An example for a non-noetherian ring is {f Q[x] : f (0) Z} (this was proved
in lectures).
Theorem 5.1. If R is noetherian, then so is R[x].
Proof. Let I be an ideal of R[x]. Our goal will be to generate I with a finite
number of elements. Let Xn be the set of polynomials in I of degree exactly n.
Then I is a disjoint union: {0} X0 . Let Jn be the following set:
Jn = {0} {r R s.t.f Xn with f = rxn + . . . }.
So we have Jn R, i.e. Jn = 0+leading terms of elements of Xn .

37

Claim. Jn is an ideal of R.
Proof. Jn is closed under since if r, s Jn r, s 6= 0, then f, g I such that
f + rxn . . . , g = sxn + . . . , and so
f g I (r s)xn + Xn if r s 6= 0
r s Jn . The cases where one of r, sr s = 0 can be verified by hand. For
example, if r 6= 0, r Jn , then to show thatr Jn ,
r Jn = f = rxn + I = f I = r Jn .
Now say that r Jn , s R. We want to show that rs Jn . For rs = 0, this is
clear. If rs 6= 0, then r 6= 0, and so
f = rxn + Xn I sf I
with
sf = srxn + . . . = sr Jn .
Claim. J0 J1 J2 . . . .
Proof. If r Jn , then either r = 0 r Jn+1 or r 6= 0 and
f = rxn + I.
x R[x] = xf I and xf = rxn+1 + Xn+1 = xf Xn+1 and
r Jn+1 .
Claim. N such thatJN = JN +1 = = JN for all n N (c.f. proof that a
PID is a UFD).
S
Proof. If J = n1 Jn , then J is an ideal (seen this before) and R is Noetherian.
Hence J is finitely-generated and j1 , . . . jm J such that J = (j1 , . . . jm ). Now,
each ji is in some J for some (and hence all) sufficiently large . Let N = max
of the . Then N is such that j1 , . . . , jm JN . So
J = (j1 , . . . jm ) JN JN +1 J
and hence all inclusions are equalities.
Now, for 0 n N, Jn is an ideal of R, and hence Jn is finitely-generated, say
by rn1 , rn2 , . . . , rnn , where n 6= the number of generators. Choose fnk Xn
such that
fnk = rnk xn + . . . , 1 k n .
{fnk } is a finite set, and fnk Xn I for all n and for all k.
Claim. The finite set {fnk } 0 n N generates I.
0 k n
Proof. We will check that f I, f (f01 , f02 , . . . , f0k , . . . ) = I 0 I by
induction on deg f . Clear for deg f = 0, i.e.f R, f 6= 0. Then
f X0 f J0 = X0 {0} = f (f01 , f02 , . . . , f00 )
38

as an ideal of R. Therefore
f

ri f0i ri R f I.

Next case: 1 deg f N, f I. We can assume any polynomial in I of degree

stricly less than deg f in I 0 . Want f I 0 . Set d = deg f . Then
f = rxd + . . . , r J .
P
By definition, Jd = (rd0 , rd1 , . . . ) and so i R such that r = i P
i fdi (think
of these as leading terms). Now consider the polynomial g = f i i fdi . f
and all the fdi have degree d. Hence deg g d. But
!
X
g= r
i rdi xd + = 0xd + . . .
i

deg g < d.
Furthermore, f and the fdi I g I. The inductive hypothesis implies that
g I 0 . Hence
X
f =g+
i fdi I 0 .
I 0

Final case: deg f = d > N. Again, can assume (inductive hypothesis).

g I, deg(g) < d = g I 0 .
d
Again, write
P f = rx + . . . where r Jd = JN = (rN 0 , rN 1 , . . . ). Hence we can
write r = i i rN i i R. Consider

g=f

i xdN

fN i
degreeN

degree d, leading term is i rN i . Now

X
g = rxd
(i rN i )xd + l.o.t.
i

therefore deg g < d g I 0 (by inductive hypothesis) and so f I 0 .

Corollary 5.2. If R is a field or a PID, then R[x1 , . . . , xn ] is noetherian.
Proof. R is noetherian. Now we use induction on n.
Lemma 5.3. If : A B is a surjective ring homomorphism, and if A is
noetherian, then B is noetherian.
Remark. If B A is a subring, then this is NOT true.
A noetherian = B noetherian.
For example A = Q[x], B =?????????

39

Proof. Say I B is an ideal of B. Set J = {a A|(a) I} = 1 (I). Then

J is an ideal of A by either checking hte axioms or by observing that J = ker
where
/
/ B/I.
B
A
;

Hence J = (a1 , a2 , . . . an ) for some finite set {ai }. Set bi = (ai ). I claim that
I = (b1 , . . . , bn ). For if b I, is surjective, we have that a A such that
(a) = b, (a) I = a J.
Hence a =

Pn

i=1

i ai i A. By applying , we get
b=

n
X

(i ) ai (b1 , . . . , bn ).

i=1

Corollary 5.4. If R is a PID and I R[x1 , . . . , xn ] is an ideal, then R[x1 , . . . , xn ]/I

is noetherian.

Example. Z[ 5] is not a UFD (and in fact not a PID: the ideal (2, 1 + 5)
turns out not to be principal).
But is is noetherian. there is a surjection
Z[x] Z[ 5] given by x 7 5

x2 + 2x 7 7 ( 5)2 + 2 5 7 etc.

In particular a + bx 7 a + b 5. Hence this map is surjective (notice that its

kernel is (x2 + 5)). As
Z is a PID, it follows that Z is noetherian. Hence Z[x] is
noetherian and so Z[ 5] is notherian by 5.1 and 5.3 respectively.
Remark. In a P15 you will study the integer ring of a number field.
Our next goal will be to show that if R is a UFD, then R[x] is also a UFD.
Say R is a UFD. Note that if we define a relation on a set of irreducible
elements by
a b a = ub, u a unit
i.e. a b = a, b are associates, then is an equivalence relation. Hence
it divides the set of irreducibles into equivalence classes. Now, choose a fixed
representative in each class and let X be the set of representatives. For example
if R = Z, p p and p p, p p for p prime. These are all, hence the
equivalence class of p is {p, p} and X = {2, 3, 5, 7, . . . } is a possible set of
representatives. Now, each non-zero element r of R can be written uniquely (up
to reordering) as r = uxe11 . . . xenn for some n 0, u a unit, xj X, ej Z 1
. Hence if F = F rac(R), then any f 6= 0 F can be written as f = rs (factors
r and s) and hence f = uxe11 . . . xenn with now ej Z. For example, in Q for
62
= 21 311 171 .
X = {2, 3, 5, . . . }, we have 12 = 22 3, 9 = 32 and 17
Definition 5.5. Fix an irreducible element X R, where R is a UFD.
The adic valuation on F = F rac(R) is the map v = val : F Z defined
as follows: if 0 6= f F, then write
Y
f = u e
xpi i
xi X,xi 6=

40

with e Z (e = 0 is possible) and define val (f ) = e. It is easily checked that

val(f g) = val(f )+val(g) (hence v is a group homomorphism (F , ) (Z, +)).
Remark. Sometimes it is a useful convention to set val (0) = + so that
val : F Z {+}, for example R = Z, X = {2, 3, . . . } and = 2. val (8) =
3, val (70) = 1, since 70 = 21 5 7. Also val ( 43 ) = 2 and val (x odd) = 0.
Definition 5.6. Say 0 6= q F [x], q = f0 + f1 x + + fn xn . Say X. Define
v (q) = min {v (fi )}.
i:fi 6=0

Example.
- Let R = C[z], and f = z3
22 . The irreducibles in R are X = {z | C}.
val2 (f ) = 2 and valz3 (f ) = 1, so valz (f ) = 0 6= 0, 3.
- Let R = Z, f =
0 p 6= 3, 7.

7
9

7
32 .

Hence val3 (f ) = 2, val7 (f ) = 1, valp (f ) =

Lemma 5.7.
(i) val (xy) = val (x)val (y)
(ii) if val (x) = 0 and val (y) > 0, then val (x + y) = 0.
Proof. (i) x = u e

Qn

i=1

xei i (it is okay if some ei = 0) and y = v d

xy = uv e+d

n
Y

Qn

j=1

xj j . Therefore

xidi +ei ,

i=1

hence (i) is true.

(ii) Let x = rs , | r, | s and y =
x+y =

e t
w ,

| t, | w, e 1. Then

rw + e ts
r e t
+
=
.
s
w
ws

Note that - s, - w - sw (as irreducibles are prime in a UFD).

Furthermore, if | (rw + e ts) then | rw | r or | w, contradiction.
Hence - (rw + e ts) = val (x + y) = 0, so (ii) is also true.
Pd
This extends to polynomials. Say 0 6= f F [x], f = i=0 ai xi . Define val (f ) =
mini val (ai ) (or val (f ) = mini:ai 6=0 val (ai )). Note that factorsation of each
of the ai only involves finitely many s and f has only finitely many ai s.
Therefore val (f ) = 0 for all but finitely many .
Definition 5.8. The content of 0 6= f F [x] is
Y
cont(f ) =
val (f ) = a finite product.
X

We say 0 6= f F [x] is primitive if cont(f ) = 1.

41

Example. Let R = Z, X = {2, 3, 5 . . . } and f (x) = 94 + 45x + 21x2 . Note that

9
32
2
4 = 22 , 45 = 3 5, 21 = 3 7. Therefore the only irreducibles that show up
are 2,3,5,7 valp (f ) = 0 p > 7. Now
p
2
3
5
7

valp ( 94 )
-2
2
0
0

valp (45)
0
2
1
0

valp (21)
0
1
0
1

valp (f )
-2
1
0
0

Hence cont(f ) = 22 3 = 34 .
Note.


f (x)/cont(f ) =


3
9
+ 45x + 21x2 /( ) = 3 + 60x + 28x2
4
4

which is a polynomial with all coefficients in Z and gcd = 1.

Remark. The content of f (x) is only defined up to a unit - for example if we
had chosen X = {2, 3, 5, . . . }, then cont(f ) = (2)2 (3)1 = 43 . So
think of cont(f ) as only defined up to a unit.
Note. In general, cont(f ) F .
When is the content in R? Well,
Q
valx (f )
cont(f ) R
R valx (f ) 0 x X
xX x

mini valx (ai ) 0 x, i valx (ai ) 0 x, i

ai R i f R[x].
Furthermore,
f primitive cont(f ) = 1 min valx (ai ) = 0 x x, i s.t. x - ai
i

f R[x] and gcdof all coefficients is 1.

Lemma 5.9. If 0 6= f F [x] and F , then
cont(f ) = ucont(f ) u a unit.
Proof. Let f =

ai xi , then

valx (f ) = min valx (ai ) = min(valx () + valx (ai ))

i

by 5.6(i). So
valx (f ) = valx () + min(valx (ai )) = valx () + valx (f )
i

and the lemma follows as

xvalx () = u, where u is a unit..

Lemma 5.10. If R is a UFD and F = F rac(R) and if f, g F [x] are both

non-zero, then
cont(f g) = cont(f )cont(g).
42

Proof. Let c = cont(f ), d = cont(g). Then f = cf1 , g = dg1 and Lemma

5.9 implies that cont(f1 ) = cont(g1 ) = 1. Again by Lemma 5.9, STP that
cont(f1 g1 ) = 1. STP cont(f
P ) = cont(g) = 1 = cont(f g) = 1, and now
cont(f ) = 1 = f =P ai xi where ai R andfor irreducible i such that
- ai . Similarly, g =
bj xj , bj R and no divides all of them. Certainly,
then, f g R[x] cont(f g) R (as opposed to F ), i.e.
cont(f ) = cont(g) = 1 = f, g R[x] = f g R[x] = cont(f g) R
(i.e. cont(h) R h R[x]). Now, we want to prove that cont(f g) = 1.
We will prove this by contradiction. Say cont(f g) 6= 1. ChoosePan irreducible
R such that | cont(f g). Now - cont(f ), and so if f = i0 ai xi , then
we know that i such that | ai . Choose
the biggest such i, so - ai but
P
| ai+1 , | ai+2 , . . . . Similarly g = j0 bj xj , and choose the biggest j such
P
that - bj and | bj+1 | bj+2 , . . . . Now say f g = k0 ck xk . As | cont(f g),
we must have | ck k. But I claim that - ci+j . For
ci+j = a0 bi+j + a1 bi+j1 + + ai1 bj+1 + ai bj + ai+1 bj1 + + ai+j b0 .
Notice that by the choice of i and j, | bn if n > j and | an if n > i. Hence
all terms in ci+j are multiples of except ai bj . So - ci+j (by Lemma 5.7 if
you like: v (ai ) = v (bj ) = 0 = v (ai bj ) = 0 and v 1 for all other terms
in ci+j ). This proves the lemma.
Corollary 5.11. (Gauss Lemma) Let R be a UFD and 0 6= f R[x] with
deg f 1. If f is irreducible in R[x], then f is irreducible in F [x].
Proof. Say f R[x], deg f 1 and say f is irreducible in F [x] and in R[x]. Let
us arrive at a contradiction.
First note that f must be primitive (i.e. cont(f ) =
P
1), because if f =
ai xi and is irreducible and | ai i, then f = g, where
, g are not units and so f is factored. Say f = gh in F [x], deg(g), deg(h) > 0.
By 5.10, if a = cont(g), b = cont(h), then ab = 1. Now set g 0 = ag , h0 = hb . Then
f = g 0 h0 and cont(g 0 ) = cont(h0 ) = 1. Therefore g 0 , h0 R[x]. Hence f is not
irreducible in R[x]. Contradiction. This proves the corollary.
in fact, one can greatly generalise 5.11 to really, fully understand factorisation
in R[x].
Proposition 5.12. (Factorisation in R[x]) Let R be a UFD and f R[x], f 6=
0. Then f can be written as f = cg1 g2 . . . gn with c R, gi R[x] where gi is
primitive and gi is irreducible in F [x], where F = F rac(R). Furthermore, this
factorisation is unique up to
(1) changing factors by units
(2) reordering the gi .
Proof. Let F be Q
a field. Then F [x] is a UFD (by 3.8 and 3.11), hence we
can factor f = u i gi , where gi F [x] is irreducible and u FQ =units in
F [x]. Set gi = gi /cont(
gi ). Then gi is primitive by 5.8 and f = c i gi (x) with
c F . Taking contents, we see that cont(f ) = cont(c) = c , a unit in R.
This proves existence. Now, for uniqueness, say f = cg1 g2 . . . gm = c0 g10 g20 . . . gn0 .
43

We know that F [x] is a UFD, hence n = m and after reordering, we have

gi0 = i gi , F . Hence
cont(gi0 ) = unit in R i cont(gi ) = i R .
=1

=1

Finally, taking contents of

0
cg1 . . . gm = c0 g10 . . . gm
= c = uxc0 , u R .

Theorem 5.13. If R is a UFD, then so is R[x]. Furthermore, the irreducible

elements of R[x] are of two kinds:
(1) The irreducible elements of R.
(2) Primitive polynomials in R[x] which are irreducible in F [x].
Moreover, the units in R[x] are the units in R (easy to show).
Proof. Let R be a UFD. Then R is an ID = R[x] is an ID (by 1.13). By
considering degrees
= (deg f > 0 = f not a unit) = R[x] = R .
The axioms UF1 and UF2 of a UFD are precisely 5.12.
Corollary 5.14. R a UFD = R[x1 , x2 , . . . , xn ] is a UFD.
Proof. Trivial.
Example. We know that Z[x] is a UFD, hence Z[x1 , . . . , xn ] is a UFD. Also, if
k is a field, then k[x1 , . . . , xn ] is a UFD.
Remark. Z[x] is not a PID (the ideal (2, x) is not principal) and k[x, y] is also
not a PID (the ideal (x, y) is not principal).

44

Chapter 6

Modules
A module can be viewed as a vector space over a ring. In this chapter, R will
denote a ring (not necessarily commutative).
Definition 6.1. An Rmodule (or more precisely a left-module), is an abelian
group M equipped with a map R M M, traditionally denoted (r, m) 7 rm
such that
1) r(m + n) = rm + rn r R, m, n M
2) (r + s)m = rm + sm r, s R, m M
3) (rs)m = r(sm) r, s R, m M
4) 1m = m m M .
Remark. If R is a field, then these are precisely the axioms for a vector space
over a field.
Remark. It is called a left Rmodule, because the action of R on M is on the
left: rm, not mr.
Remark. If R is non-commutative, the notions of a left Rmodule and a right
Rmodule do not concide. If R is commutative and M is a left Rmodule, then
we can make M into a right Rmodule by defining mr := rm and checking that
the axioms are still met.
Exercise. Let R be a non-commutative ring and M be a left Rmodule. Why
does the trick mentioned above not make M into a right Rmodule?
Basic consequences of the axioms: (rs)m = r(sm). Set r = 1 and obtain
(s)m = (1)(sm) and (1)x = x because
(1)x + x = (1)x + (1)x = (1 + 1)x = 0x = 0
because
(1 + 0)x = x = 1x + 0x = x + 0x.
In other words, 0x = 0 for all x M and (1)x = x. Hence 0 is the identity
for M.
45

Example.
1. If R is a field, an Rmodule is a vector space over R.
2. R is naturally an Rmodule: R is an abelian group, and define a map
R R R by multiplication in R. Then axioms 1. and 2. follow from
distributivity in R and 3. and 4. are the axioms for multiplication. More
generally, set M = Rn = column vectors in R. Then

r1
1
2 r2

r . = .
.. ..
rn
n
makes Rn into an Rmodule.
3. Let R be a ring and I be a left ideal. By definition, a left ideal is an
abelian group and if r R and i I, then ri I. Define a map R I I
by (r, i) 7 ri. This makes I into an Rmodule (all four axioms follow as
above, in the case where I = R).
4. R = Z. Say M is an abelian group. We will turn M into a Zmodule.
For m M , define
- 0m = 0
- 1m = m
- 2 m = (1 + 1)m = m + m
RM

- if t Z0 , tm = m + m + + m and (t)m = (tm).

t times
These definitions were forced upon us by the axioms and it is a relatively
easy check to show that the axioms are satisfied. Conclusion: If M is
an abelian group, then M can be given the structure of a Zmodule
in a unique way. More vaguely: abelian groups = Zmodules. Now
we clearly see that it is not the case that a general (finitely-generated)
Rmodule is isomorphic to Rn for some n. For example, Z/nZ is a
Zmodule. Clearly not Zm for any m.
5. Let F be a field, R = Mn (F ) = set of n n matrices over F . Let
M = F n = column vectors of length n. Then R acts on M in an obvious
way. We check the axioms and conclude that M is an Rmodule.
6. Let F be a field and let R = F [x]. Let M be any vector space over F . M
is an F -module, but it does not have an action yet. Let : M M be
any F linear map. Define (m) = xm. More generally, define
xn m = ((. . . (m) . . . )) = n m
n times
and if f =

ai xi F [x], define
fm =

X
i

46

ai i m.
F M

This makes M into an F [x]module. F [x]modules = F - vector

spaces equipped with a linear map .
7. Recall that a representation of a group G is simply a group homomorphism
: G GL(V ) where V is a finite dimensional vector space over, say, C.
If R is the group ring of G, i.e. R = Cg1 Cg2 Cgn is a vector
space with basis G, where G = {g1 , . . . , gn } and g h = gh and extends
linearly. Then R becomes a ring and V becomes an Rmodule.
Definition 6.2. Let R be a ring and M1 and M2 be Rmodules. An Rmodule
homomorphism f : M1 M2 is a homomorphism of abelian groups f : M1
M2 such that f (rm1 ) = rf (m1 ). An endomorphism of an Rmodule M is an
Rmodule homomorphism f : M M . An isomorphism of Rmodules is
an Rmodule homomorphism f : M1 M2 such that there exists g : M2
M1 also an Rmodule homomorphism such that f g = g f = identity.
Equivalently, an Rmodule homomorphism is an isomorphism if it is bijective.
An automorphism is a bijective endomorphism.
Example.
1. Let R be a field and V, W be Rmodules (i.e. vector spaces) then an
Rmodule homomorphism V W is just a linear map V W.
2. R = Z. An Rmodule is an abelian group and an Rmodule homomorphism f : A B is an abelian group homomorphism (because, e.g.
f (3a) = 3f (a) automatically as f (3a) = f (a+a+a) = f (a)+f (a)+f (a) =
3f (a)).
3. (obscure) If f : A B is a ring homomorphism, then B becomes an
Amodule. Thus:
a b = f (a)b B
(easy exercise: show that the axioms are met and f is an Amodule
homomorphism).
4. Let R be a commutative ring and M1 = Rn and M2 = Rm . Let X be
the m n matrix with entries in R. Then X induces an Rmodule
homomorphism M1 M2 in an obvious way: XV is the result of matrix
multiplication.
Definition 6.3. Let M be an Rmodule. A subgroup P of M is an Rsubmodule
or just a submodule, if r R, p P, rp P.
Example.
1. Let R be a field = M a vector space and P is just a subspace.
2. Let R be a ring and M = R. A submodule of R is a subgroup I which
is closed under left multiplication by R, i.e. r R, i I = ri I.
Therefore, a submodule of R is a left ideal of R.
3. Say that R is a ring and M1 , M2 are Rmodules and f : M1 M2
is an Rmodule homomorphism. The kernel of f (as an abelian group

47

homomorphism) is a submodule of M1 if the image of f is a submodule of

M2 .
Proof. Set K = ker f = {m1 M1 : f (m1 ) = 0}. By group theory, we
know that K is an abelian subgroup. Say k K and r R, then
f (rk) = rf (k) = r0 = 0 = rk K
so K is a submodule. Now let I = Imf = {f (m1 ) : m1 M1 } M2 . By
group theory, I is a subgroup. Now say i I and r R. Then by definition
i = f (m1 ) for some m1 M1 and note that f (rm1 ) = rf (m1 ) = ri which
is in I. Hence ri Iand I is indeed a submodule.
Quotient Modules
Let R be a ring, M an Rmodule and P M be a Rsubmodule. By group
theory, we know that the quotient M/P is an abelian group. Now, we want to
give M/P the structure of an Rmodule. A general element of M/P is of the
form m + P. Define
r(m + P ) = rm + P.
Is this well-defined? Well, say m + P = m0 + P, i.e. m0 = m + p for some p P .
To be well-defined, we need rm + P = rm0 + P. But this is true, as
rm0 rm = rm + rp rm = rp
and P is a submodule: rp P = rm + P = rm0 + P. So M/P has a
natural Raction. From here, it is easy to check the axioms.
Definition 6.4. The quotient module M/P is an abelian group with the above
action (1st isomorphism for vector spaces (?) : V W, V / ker
= Im.
Therefore, dim V dim(ker ) = dim(Im)).
Proposition 6.5. Let R be a ring (not necessary, actually) and f : M N
an Rmodule homomorphism. Then the natural map from M/ ker f to Imf is
an isomorphism of Rmodules.
Proof. f is an abelian group homomorphism. By the first isomorphism theorem
for groups, there exists a natural map M/ ker f Imf given by m + ker f 7
f (m). This map is an isomorphism of abelian groups. It is only left to check
that this is an Rmodule homomorphism, i.e. STP that r R, m M,
r(m + ker f ) rf (m), i.e. STP f (rm) = rf (m). But this is true by definition
of an Rmodule.
Example. The abelian group {0} is an Rmodule. r0 = 0 for all r. It is
R/I if I = R. Further, N, ! an Rmodule homomorphism 0 N given by
0 7 0 and M, ! Rmodule homomorphism M 0 given by m 7 0 m. The
composition of these maps is the zero map 0 : M N such that m 7 0 for all
m.
Definition 6.6. If M, N are Rmodules, their direct sum M N is the abelian
group M N with R acting diagonally, i.e.
Lnr(m, n) = (rm, rn). More generally,
if Mi are modules for 1 i n, then
i=1 Mi is too: r(m1 , m2 , . . . , mn ) =
(rm1 , rm2 , . . . , rmn ).
48

We have already seen an example for this: Rn = R R.

Definition 6.7. A module M for a ring R is finitely generated if m1 , . . . , mn
M such that any m M can be written as
m = r1 m1 + + rn mn .
Remark. {r1 m1 + +rn mn : ri R} is checked to be the smallest Rsubmodule
of M containing {m1 , . . . , mn }. Hence m1 , . . . , mn generate M the smallest submodule of M containing m1 , . . . , mn is M .
Remark. As opposed to the study of vector spaces, bases are much rarer in the
theory of modules - most modules are not free, i.e. have no basis.
Remark. If R is a commutative ring and I is an ideal, then I is f.g. as an ideal
I is f.g. as Rmodule.
Lemma 6.8. A module M is finitely-generated n 0 and a surjection
: Rn M of Rmodules (i.e. is an Rmodule homomorphism).
Remark. If R is a field, an Rmodule V is f g dim V < , and then
V
= Rn for some n. Lemma 6.8. is some weak analogy of this for general R.
Proof.
( = ) Say M is generated
by m1 , . . . , mn . Define : Rn M by
P
(r1 , . . . , rn ) = ri mi . It is easy to check that is an Rmodule homomorphism. By the definition of finitely generated, is a surjection.
(=) Say : Rn M is a surjection of Rmodules. Set
mi = (0, 0, . . .

ith

1
, 0, . . . 0).
place

I claim that the mi for 1 i n generate M. For if m M, is

a surjection. Therefore there exist r1 , . . . , rn such that (r1 ,P
. . . , rn ) =
m. But if ei = (0, 0, . . .
1
, 0, . . . 0), then (r1 , . . . , rn ) =
ri ei and
ith place
applying , we get
X
X
m=
ri (ei ) =
ri mi
where we used the fact that is a Rmodule homomorphism.
Remark. Q is not a f.g. Zmodule. For if m1 , . . . , mn Q, mi = dcii with
ci , di P
Z, di 6= 0, and if D = gcd(di ), then each mi is of the form aDi , ai Z and
hence
ri mi is also of this form if ri R = Z. Therefore
X
1
6=
ri mi , ri Z.
2D
Therefore Q is not finitely generated as Zmodule.
Definition 6.9. Say M1 , . . . , Mn are Rmodules and if fi : Mi Mi+1 are
Rmodule homomorphisms, 1 i n 1. We say that this collection of
modules and maps is a complex if fi fi1 = 0 for 2 i n 1. Equivalently,
it is a complex iff
49

Imfi1 ker fi

2 i n 1.

The image of fi1 and the kernel of fi are both submodules of M.

M1

f1

/ M2
O

/ M3
O

f2

need Imf1 ker f2

/ ...

f3

fn2

/ Mn1
O

Imfn2 ker fn1

Imf2 ker f2

The definition needs to be checked at Mi iff whenever Mi has an arrow in and

an arrow out (definition continued). We say that this collection of modules and
maps is an exact sequence if Im(fi1 ) = ker(fi ) for 2 i n 1. A morphism
of exact sequences (i.e. a homomorphism of exact sequences) from
M1

/ M2

/ ...

/ Mn

N1

/ N2

/ ...

/ Nn

to

is a collection of maps fi : Mi Ni such that the resulting diagram

M1

/ M2

d1

f1


N1

d2

e1

dn1

fn1

f2


/ N2

/ ...

e2

/ Mn
fn

/ . . .

e3


/ Nn

commutes.
To check that the diagram commutes it is necessary and sufficient to check that
each square
Mi

di

fi


Ni

ei

/ Mi+1


fi+1

/ Ni+1

commutes for 1 i n 1. The same definition (for commutative) holds for

complexes in general.
Remark. If
0
and
0

/A

/B

/C

/ A0

/ B0

/ C0

/0
/0

are two short exact sequences, then a map between them is a collection of three
maps: A A0 , B B 0 , C C 0 as there is a unique map 0 0 and similarly,
we only need to check that the two middle squares commute. Again, something
that has to be checked at Mi Mi has an arrow in and an arrow out.
50

f1

/ Mn

Note. One also sees examples where the sequence goes off to infinity in either,
or both, directions. We can make an (obvious) modification to the definition:
/ M2

M1

/ M2

/ ...

/ Mn

/ ...

/ M1

/ M0

/ ...

Example.
1. A

/B

/ 0 with B 0 the zero map. This sequence is exact iff

Imf = ker(zero map) Imf = B f is injective.

2. 0

/A

/ B is exact iff

ker f = Im(zero map) ker f = 0 f is injective.

/A f /B
/ 0 is exact iff it is exact at A and B f is
3. 0
injective and surjective f is an isomorphism.
The next example is so ubiquitous, it has its own special name.
Definition 6.10. An exact sequence of the following form
0

/A

/B

/C

/0

is called a short exact sequence. Alternatively, it is called an extension of C by

A.
Definition 6.11. A commutative diagram of R-modules and R-module maps
is a collection of R-modules and maps between them such that any two induced
maps between any two R-modules in the diagram coincide.
What does it actually mean for a complex
0

/A

/B

/C

/0

to be a short exact sequence?

Well, exactness at A implies that Im 0 = ker f , i.e. f is injective. Exactness at
C implies that g is surjective. Exactness at B says that Imf = ker g.
Say we only know g : B C. Can we reconstruct f and A?
The answer is yes. f is injective and hence identifies A with ker g = Imf . Hence
given g, the only way (up to isomorphism) of building a short exact sequence
with g : B C is
0

/ ker Inclusion
/B
g
51

/C

/0

(up to isomorphism). Similarly, if

/A

/B

/C

/0

is exact and we know f , A and B, then the sequence is isomorphic to

/A

/B

/0

Imf

i.e. C must be isomorphic to B/Imf .

Question: Given only A and C, can we reconstruct B?
We need to find B which contains A and maps to C. Here is an idea that works:
Define B = A C and define f : A B by f (a) = (a, 0). Then this is clearly an
injection (and an R-module homomorphism). Define g : B C by g(a, c) = c,
i.e. the projection. Clearly a surjection. Also it is an R-module homomorphism.
Now
ker{(a, c) : c = 0} = {(a, 0) : a A} = Imf.
Is this the only possibility for B? No!
For example, R = Z, so R-module = abelian group and R-module homomorphism =
abelian group homomorphism. Consider the following short exact sequence:
/

Z
2Z

f:

0
1

where

and
0
1
g:
2
3

Z
4Z

/0

Z
2Z

7 0
7 2
7
7

0
1
.
0
1

It is clear that f is injective and that g is surjective. Also f, g are homomorphisms and Imf = {0, 2} = ker g, so this is a short exact sequence.
Is this exact sequence isomorphic to
/

0
No, as

Z
4Z

Z
2Z

Z
2Z

Z
2Z

Z
2Z

/ 0?

Z
2Z

Z
2Z .

Theorem 6.12. Say R is a ring and

0

/A

/B

/C

/0

is a short exact sequence. Then the following are equivalent:

(i) There exists : C B ( for section) a R-module homomorphism such
that g = Id : C C.

52

(ii) There exist an isomorphism of short exact sequences

/A

/C

Id


/A

/B

f0


/ AC

/0

Id


/C

g0

/0

where f 0 : A A C is given by a 7 (a, 0) and g 0 : A C C by

(a, c) 7 c such that is an isomorphism.
Proof. Assume that our short exact sequence is isomorphic to
0

/A

f0

/ AC

g0

/C

/0.

Define : C A C such that (c) = (0, c). Then

(g 0 )(c) = g 0 ((c)) = g(0, c) = c.
For the converse, say that : C B is an R-module homomorphism such that
g is the identity. Define : A C B by
(a, c) = f (a) + (c).
It is easy to see that is an R-module homomorphism. I claim that is even
an isomorphism. To see injectivity, say that (a, c) = 0. Then f (a) + (c) = 0.
Apply g. Then
g f (a) + g (c) = 0.
But g f = 0 (as this is a complex) and g (c) = c by the definition of . Hence
c = 0 = (c) = 0, and therefore f (a) = 0. But f is injective, hence a = 0.
Therefore is injective. For surjectivity, say b B. Set c = g(b). Consider
b (c) B. This element is in the kernel of g as
g(b (c)) = g(b) g (c) = c c = 0.
Hence b (c) Im(f ). Say b (c) = f (a). Then
b = f (a) + (c) = (a, c).
Hence surjective. Therefore is an isomorphism. The only thing left to show is
that the diagram commutes. There are two squares to check. Both are fine. To
complete the proof set = 1 .
Definition 6.13. A short exact sequence is called split if equivalence condition
of 6.12 holds.
Remark. 0 Z/2 Z/4 Z/2 0 is an example of a non-split s.e.s. Also,
given A, C, one can put a group structure on the iso. classes of short exact
sequences 0 A B C 0. B = A C : identity element of this group,
this group is called ExtR (C, A).

53

Definition 6.14. If f : M N is an Rmodule homomorphism, then the

cokernel of f , denoted coker(f ), is N/f (M ) = N/Im(f ).
Remark. If f : M N, then it is easy to check that the following sequence is
exact:
f

0 ker(f ) M N coker(f ) 0
and this is the reason for the notation.
Snake Lemma
Set-up: Let R be a ring and consider two short exact sequences of Rmodules
0ABC0

0 A0 B 0 C 0 0
and a given map between these sequences:
0

/A


/ A0

/B


/ B0

/0

f0

/C

g0


/ C0

/0

such that f = f 0 and g = g 0 . Let me now, given this set up,

construct a map : ker() coker().
Construction: say c ker(). g is surjective b such that g(b) = c. Now
(commutativity of right hand square)
(g(b)) = (c) = 0 g 0 ((b)) = 0 = (b) ker(g 0 ) = Im(f 0 )
= (b) = f 0 (a0 )
for some a0 A0 . Now, let
(c) = Image of a0 in A0 /A=coker().
Claim. is a well-defined Rmodule homomorphism.
To show this, say b B and g(b) = c. Then
b b ker(g) = Im(f ) b b = f (
a), a
A.
Recall (b) = f 0 (a0 ); similarly (b) = f 0 (
a0 ). The well-definedness of is pre0
0
cisely the statement that a and a
have the same image in A0 /(A), i.e. we
0
0
need to check a a
Im(). But a0 a
0 = (
a). So is well-defined. It is
easy to check that is an Rmodulo homomorphism.

54

Lemma 6.15. (Snake Lemma). With the above set-up

0

/ ker()

/ ker()

/ ker()
{
{
{{
{
{{



/A
/ B {{{{
/C
{
{
{{

{
{
{ 
{


{{ / 0
/ A0
/ C0
{{ B
{
{
{
{
{{
 }{{


/ coker()
/ coker()
coker()

/0
/0
/0

the sequence in green is exact. I.e.

0 ker() ker() ker() coker() coker() coker() 0

is exact.
Proof. (of some bits).
ker()

/A

/ ker()


/B

If a ker(), then f (a) must be in ker(), as(a) = 0 = f 0 ((a)) = 0 =

(f 0 (a)) = 0. Furthermore, it is an injection as ker(f : ker() ker()) =
ker(f : A B) ker() = 0 ker() = 0.
Exactness of ker()?
ker

/ ker

/ ker
9

need this to be zero

g : ker() C and (RH square commutes) = g(ker()) ker(). ker()
ker() ker(). Need this to be 0. But g f = 0 on all ofA. f (ker()
g
ker(ker() ker()). Finally, say b ker() and g(b) = 0. Need to check that
b = f (a) for some a ker(). But g(b) = 0 b = f (a) for some a A by
exactness of A B C. Furthermore,
b ker() (f (a)) = 0 = f 0 ((a)) = 0.
But f 0 is injective (a) = 0 exactness at ker().
Exactness at ker() : First, need to check Im(ker() ker() is contained in
ker(). So say c ker() and c = g(b), b ker(). We can use this b in definition
of , then (using that notation) b0 = (b) = 0 can set a0 = 0 and (c) = Image
of a0 in coker() = 0.
55

Finally, need to check that if c ker() and (c) = 0, then b0 ker() such
that g(b0 ) = c. Now, by definition of : c C b B such that g(b) = c.
Now, (b) Im(f 0 ) (b) = f 0 (a0 ). Assertion (c) = 0 is exactly assertion that
a0 = 0 in coker() i.e. that a0 = (a), some A. Set b0 = b f (a). Now,
g(b0 ) = g(b) gf (a) = g(b) = c. So it suffices to check that b0 ker(). Now,
(b0 ) = (b) f (a) = (b) f 0 (a) = f 0 (a0 ) f 0 (a0 ) = 0
Finally, exactness at ker() is proved.
Exercise. Do the rest!!!
Definition 6.16. Let R be a field, call it k. Let V, W be two Rmodules,
i.e. vector spaces. Let Homk (V, W ) = { : V W linear transformations}.
Observe that Homk (V, W ) is also a kvector space. Abelian group:
0=V 0W
+ : ( + )(v) = (v) + (v). If k and : V W , define : V W
by ()(v) = ((v)) = ((v)). It is easy to check that Homk (V, W ) is a vector
space.
Remark. If dimk V, dimk W < , then dimk (Homk (V, W )) = dimk (V )dimk (W ).
For say dim V = n, dim W = m. Pick bases for both. Then Homk (V, W ) becomes identified with the set of m n matrices - clearly mn dimensional.
Definition 6.17. If V is a vector space over k, the dual space V is Homk (V, k).
By the remarks above, if dim V = n < , then dim V = n 1 = n.
Lemma 6.18. Say V, W are vector spaces and : V W is a linear map.
Define : W V thus: say f W . Then f : W k f : V W k
is linear. Therefore, f V . Define f = f . Need to check that is
linear:
This is easy: For example, need (f +g) = f + g, i.e. need to check v V,
( (f + g))(v) = ( f )(v) + ( g) (v), i.e. need (f + g)(v) = f (v) + g(v).
This is true by definition of f + g etc.
is called the dual linear map.
Example. If V = k n , W = k m and : V W is represented by a matrix A
(m n), then let us compute the matrix of . Need bases for W and V . Say
e1 , . . . , en V is the usual basis, ei being the ith unit vector. Define fi V
by fi (a1 , . . . , an )> = ai . This is clearly linear, therefore fi V . Note also that
fi (ej ) = ij .
Claim. The fi are a basis for V , fi is a dual basis.
Proof. There is a right number
it suffices
to check linear indeP of them, hence
P
i fi is the zero function
pendence. SoP
say i k and
i fi = 0 P
in V , i.e.
V k j, i (i fi )(ej ) = 0 = 0 = i i ij = j j. Hence fi are linearly
indemepnt. So the fi form a basis.
Now say V is a vector space V is it dual. What is (V ) ?

56

Lemma 6.19. There is a canocial linear map V V . If dim V < , this is

an isomorphism.
Proof. (V ) = Homk (V , k). Say v V. Need (v) V , i.e. need (v) :
V k, i.e. f V , need (v)(f ) k. But f : V k. So let us define
((v))(f ) = f (v). Check
1. (v) is linear, i.e. (v) V .
2. is linear (good exercise!).
Finally, need to check is an isomorphism if dim V < . Certainly dim V =
dim v suffices to check that is injective. But say 0 6= v V and v = 0 and
let us go for a contradiction.
v = 0 means f V , (v)(f ) = 0, i.e. f (v) = 0. So to get a contraciton, need
f

a linear map V k such that f (v)

P 6= 0. But 0 6= v can extend {v} to a basis
v = e1 , e2 , . . . , en of V . Define f ( i ei ) = 1 . Linear and f (v) = 1.
Remark. If dim V < , hten V
= V , but there is no natural isomorphism; no
analogue of .

57