Você está na página 1de 150

ELEMENTARY

COMMUTATIVE ALGEBRA
LECTURE NOTES
H.A. NIELSEN
DEPARTMENT OF MATHEMATICAL SCIENCES
UNIVERSITY OF AARHUS
2005
Elementary Commutative Algebra
H.A. Nielsen
Contents
Prerequisites 7
1. A dictionary on rings and ideals 9
1.1. Rings 9
1.2. Ideals 11
1.3. Prime ideals 13
1.4. Chinese remainders 14
1.5. Unique factorization 15
1.6. Polynomials 16
1.7. Roots 18
1.8. Fields 19
1.9. Power series 20
2. Modules 21
2.1. Modules and homomorphisms 21
2.2. Submodules and factor modules 23
2.3. Kernel and cokernel 25
2.4. Sum and product 28
2.5. Homomorphism modules 30
2.6. Tensor product modules 34
2.7. Change of rings 37
3. Exact sequences of modules 41
3.1. Exact sequences 41
3.2. The snake lemma 45
3.3. Exactness of Hom 50
3.4. Exactness of tensor 53
3.5. Projective modules 55
3.6. Injective modules 56
3.7. Flat modules 60
4. Fraction constructions 63
4.1. Rings of fractions 63
4.2. Modules of fractions 65
4.3. Exactness of fractions 67
4.4. Tensor modules of fractions 69
4.5. Homomorphism modules of fractions 70
4.6. The polynomial ring is factorial 71
5. Localization 73
5.1. Prime ideals 73
5.2. Localization of rings 75
5
6 CONTENTS
5.3. Localization of modules 77
5.4. The local-global principle 79
5.5. Flat ring homomorphisms 81
5.6. Faithfully at ring homomorphisms 83
6. Finite modules 87
6.1. Finite modules 87
6.2. Free modules 89
6.3. Cayley-Hamiltons theorem 91
6.4. Nakayamas lemma 93
6.5. Finite presented modules 94
6.6. Finite ring homomorphisms 99
7. Modules of nite length 101
7.1. Simple modules 101
7.2. The length 102
7.3. Artinian modules 105
7.4. Artinian rings 107
7.5. Localization 109
7.6. Local artinian ring 110
8. Noetherian rings 113
8.1. Noetherian modules 113
8.2. Noetherian rings 115
8.3. Finite type rings 117
8.4. Power series rings 118
8.5. Localization of noetherian rings 120
8.6. Prime ltrations of modules 121
9. Primary decomposition 123
9.1. Zariski topology 123
9.2. Support of modules 126
9.3. Ass of modules 128
9.4. Primary modules 132
9.5. Decomposition of modules 134
9.6. Decomposition of ideals 136
10. Dedekind rings 139
10.1. Principal ideal domains 139
10.2. Discrete valuation rings 141
10.3. Dedekind domains 143
Bibliography 145
Index 147
Prerequisites
The basic notions from algebra, such as groups, rings, elds and their homomor-
phisms together with some linear algebra, bilinear forms, matrices and determi-
nants.
Linear algebra: Fraleigh & Beauregard, Linear algebra, New York 1995.
Algebra: Niels Lauritzen, Concrete abstract algebra, Cambridge 2003.
Also recommended: Jens Carsten Jantzen, Algebra 2, Aarhus 2004.
The propositions are stated complete and precise, while the proofs are quite short.
No specic references to the literature are given. But lacking details may all be
found at appropriate places in the books listed in the bibliography.
A proposition being important when working in commutative algebra or a propo-
sition containing a nal result is named Theorem.
Nielsen, University of Aarhus, Spring 2005
7
1
A dictionary on rings and ideals
1.1. Rings
1.1.1. Denition. An abelian group is a set Awith an addition AA A, (a, b)
a +b and a zero 0 A satisfying
(1) associative: (a +b) +c = a + (b +c)
(2) zero: a + 0 = a = 0 +a
(3) negative: a + (a) = 0
(4) commutative: a +b = b +a
for all a, b, c A. A subset B A is a subgroup if 0 B and a b B for
all a, b B. The factor group A/B is the abelian group whose elements are the
cosets a +B = a +b[b B with addition (a +B) +(b +B) = (a +b) +B. A
homomorphism of groups : A C respects addition (a + b) = (a) + (b).
The projection : A A/B, a a +B is a homomorphism. If (b) = 0 for all
b B, then there is a unique homomorphism

: A/B C such that =

.
1.1.2. Denition. A ring is an abelian group R, addition (a, b) a + b and zero
0, together with a multiplication R R R, (a, b) ab and an identity 1 R
satisfying
(1) associative: (ab)c = a(bc)
(2) distributive: a(b +c) = ab +ac, (a +b)c = ac +bc
(3) identity : 1a = a = a1
(4) commutative : ab = ba
for all a, b, c R. If (4) is not satised then R is a noncommutative ring. A
subring R

R is an additive subgroup such that 1 R

and ab R

for all
a, b R

. The inclusion R

R is a ring extension. A homomorphism of rings


: R S is an additive group homomorphism respecting multiplication and
identity
(a +b) = (a) +(b), (ab) = (a)(b), (1) = 1
An isomorphism is a homomorphism : R S having an inverse map
1
:
S R which is also a homomorphism. The identity isomorphism is denoted
1
R
: R R.
1.1.3. Remark. (1) A bijective ring homomorphism is an isomorphism.
(2) Recall the usual formulas: a + (b) = a b, 0a = 0, (1)a = a.
(3) The identity 1 is unique.
(4) A ring R is nonzero if and only if the elements 0 ,= 1.
(5) If : R S is a ring homomorphism, then (0) = 0 and (R) is a subring
of S.
(6) The unique additive group homomorphism Z R, 1 1 is a ring homo-
morphism.
9
10 1. A DICTIONARY ON RINGS AND IDEALS
1.1.4. Proposition. Let R
1
, R
2
be rings. The product ring is the product of addi-
tive groups R
1
R
2
, ((a
1
, a
2
), (b
1
, b
2
)) (a
1
+b
1
, a
2
+b
2
), with coordinate multi-
plication ((a
1
, a
2
), (b
1
, b
2
)) (a
1
b
1
, a
2
b
2
). The element (1, 1) is the identity. The
projections R
1
R
2
R
1
, (a
1
, a
2
) a
1
and R
1
R
2
R
2
, (a
1
, a
2
) a
2
are
ring homomorphisms.
1.1.5. Lemma. In a ring R the binomial formula is true
(a +b)
n
=
n

k=0
_
n
k
_
a
nk
b
k
a, b R and n a positive integer.
Proof. The multiplication is commutative, so the usual proof for numbers works.
Use the binomial identity
_
n
k 1
_
+
_
n
k
_
=
_
n + 1
k
_
together with induction on n.
1.1.6. Denition. a R is a nonzero divisor if ab ,= 0 for all b ,= 0 otherwise a
zero divisor. a is a unit if there is a b such that ab = 1.
1.1.7. Remark. (1) A unit is a nonzero divisor.
(2) If ab = 1 then b is uniquely determined by a and denoted b = a
1
.
1.1.8. Denition. A nonzero ring R is a domain if every nonzero element is a
nonzero divisor and a eld if every nonzero element is a unit. Clearly a eld is a
domain.
1.1.9. Example. The integers Zis a domain. The units in Zare 1. The rational
numbers Q, the real numbers R and the complex numbers C are elds. The
natural numbers Nis not a ring.
1.1.10. Example. The set of n n-matrices with entries from a commutative ring
is an important normally noncommutative ring.
1.1.11. Exercise. (1) Show that the product of two domains is never a domain.
(2) Let R be a ring. Show that the set of matrices
U
2
=
__
a b
0 a
_

a, b R
_
with matrix addition and matrix multiplication is a ring.
(3) Show that the set of matrices with real number entries
__
a b
b a
_

a, b R
_
with matrix addition and multiplication is a eld isomorphic to C.
(4) Show that the composition of two ring homomorphisms is again a ring homomor-
phism.
(5) Show the claim 1.1.3 that a bijective ring homomorphism is a ring isomorphism.
(6) Let : 0 R be a ring homomorphism from the zero ring. Show that R is itself the
zero ring.
1.2. IDEALS 11
1.2. Ideals
1.2.1. Denition. Let R be a ring. An ideal I is an additive subgroup of R such
that ab I for all a R, b I. A proper ideal is an ideal I ,= R.
1.2.2. Lemma. Let I

be a family of ideals in R. Then the additive subgroups

and

are ideals.
Proof. The claim for the intersection is clear. Use the formulas

(b

+c

) and a

ab

to conclude the claim for the sum.


1.2.3. Denition. The intersection, 1.2.2, of all ideals containing a subset B R
is the ideal generated by B and denoted (B) = RB = BR. It is the smallest ideal
containing B. A principal ideal (b) = Rb is an ideal generated by one element. A
nite ideal (b
1
, . . . , b
n
) is an ideal generated by nitely many elements. The zero
ideal is (0) = 0. The ring itself is a principal ideal, (1) = R. The ideal product
of two ideals I, J is denoted IJ and is the ideal generated by all ab, a I, b J.
This generalizes to the product of nitely many ideals. The power of an ideal is
denoted I
n
. The colon ideal I : J is the ideal of elements a R such that aJ I.
1.2.4. Example. (1) Every ideal in Z is principal.
(2) In a eld (0), (1) are the only ideals.
(3) A subring is normally not an ideal.
(4) Let K be a eld. In K K there are four ideals (0), (1), ((1, 0)), ((0, 1)).
1.2.5. Proposition. Let R be a ring and B a subset, then RB =

bB
Rb
RB = a
1
b
1
+ +a
n
b
n
[a
i
R, b
i
B
A principal ideal is
(b) = Rb = ab[a R
A nite ideal is
(b
1
, . . . , b
n
) = Rb
1
+ +Rb
n
Proof. The righthand side is contained in the ideal RB. Moreover the righthand
side is an ideal containing B, so equality.
1.2.6. Denition. Let : R S be a ring homomorphism. For an ideal J S
the contracted ideal is
1
(J) R and denoted J R. The kernel is the ideal
Ker =
1
(0). For an ideal I R the extended ideal is the ideal (I)S S
and denoted IS. Note that (J R)S J and I (IS) R.
1.2.7. Lemma. Let I R be an ideal and let R/I be the additive factor group.
The multiplication
R/I R/I R/I, (a +I, b +I) ab +I
is well dened. Together with the addition the conditions of 1.1.2 are satised.
Proof. If a + I = a

+ I and b + I = b

+ I then a a

, b b

I and so
ab a

= a(b b

) + b

(a a

) I. Therefore ab + I = a

+ I and the
multiplication is well dened. Clearly 1.1.2 are satised.
1.2.8. Denition. Let Rbe a ring and I an ideal, then the factor ring is the additive
factor group R/I with addition (a + I, b + I) (a + b) + I and multiplication,
1.2.7, (a +I, b +I) ab +I. The projection : R R/I, a a +I is a ring
homomorphism.
12 1. A DICTIONARY ON RINGS AND IDEALS
1.2.9. Theorem. Let : R S be a ring homomorphism.
(1) Let I Ker be an ideal. Then there is a unique ring homomorphism

: R/I S such that =

.
R

S
R/I

(2) The homomorphism

: R/ Ker S is a ring isomorphism onto the


subring (R) of S.
R

(R)
R/ Ker

(3) For any ideal J S, I =


1
(J) R is an ideal and the map

: R/I
S/J is an injective ring homomorphism.
Proof. The statements are clear for the addition and the factor map

(a + I) =
(a) is clearly a ring homomorphism.
1.2.10. Corollary. Let : R R/I be the projection. The map I

J =

1
(I

) gives a bijective correspondence between ideals I

in R/I and ideals J in


Rcontaining I. Also I

= (J) = J/I. This correspondence preserves inclusions,


sums and intersections of ideals.
1.2.11. Corollary. Let I J R be ideals. Then there is a canonical isomor-
phism
R/J (R/I)/(J/I)
Proof. The kernel of the surjective east-south composite
R

R/I

R/J

(R/I)/(J/I)
is J. By 1.2.9 the horizontal lower factor map gives the isomorphism.
1.2.12. Example. For any integer n the ideals in the factor ring Z/(n) correspond
to ideals (m) Z where m divides n.
1.2.13. Denition. Let R be a ring. The additive kernel of the unique ring homo-
morphism Z R is a principal ideal generated by a natural number char(R), the
characteristic of R. Z/(char(R)) is isomorphic to the smallest subring of R.
1.2.14. Proposition. If the characteristic char(R) = p is a prime number, then
the Frobenius homomorphism R R, a a
p
is a ring homomorphism.
Proof. By the binomial formula 1.1.4
(a +b)
p
=
p

k=0
_
p
k
_
a
pk
b
k
= a
p
+b
p
1.3. PRIME IDEALS 13
since a prime number p divides
_
p
k
_
, 0 < k < p. Clearly (ab)
p
= a
p
b
p
.
1.2.15. Exercise. (1) Let I, J be ideals in R. Show that the ideal product
IJ = a
1
b
1
+ +a
n
b
n
[a
i
I, b
i
J
(2) Let I R be an ideal. Show that I = I : R.
(3) Show that a R is a unit if and only if (a) = R.
(4) Show that a ring is a eld if and only if (0) ,= (1) are the only two ideals.
(5) Show that a nonzero ring K is a eld if and only if any nonzero ring homomorphism
: K R is injective.
(6) Let m, n be natural numbers. Determine the ideals in Z
(m, n), (m) + (n), (m) (n), (m)(n)
as principal ideals.
(7) Show that a additive cyclic group has a unique ring structure.
(8) Let p be a prime number. What is the Frobenius homomorphism on the ring Z/(p)?
1.3. Prime ideals
1.3.1. Denition. Let R be a ring and P ,= R a proper ideal.
(1) P is a prime ideal if for any product ab P either a P or b P. This
amounts to: if a, b / P then ab / P.
(2) P is a maximal ideal if no proper ideal ,= P contains P.
1.3.2. Proposition. Let P be a prime ideal and I
1
, . . . I
n
ideals such that I
1
. . . I
n

P, then some I
k
P.
Proof. If there exist a
k
I
k
P for all k, then since P is prime a
1
. . . a
n

I
1
. . . I
n
P contradicting the inclusion I
1
. . . I
n
P.
1.3.3. Proposition. Let R be a ring and P an ideal.
(1) P is a prime ideal if and only if R/P is a domain.
(2) P is a maximal ideal if and only if R/P is a eld.
Proof. Remark P ,= R R/P ,= 0. (1) Assume a + P, b + P are nonzero in
R/P. Then a, b / P. So if P is prime then by 1.3.1 ab / P and ab +P is nonzero
in R/P. It follows that R/P is a domain. The converse is similar.
(2) Assume R/P is a eld and a / P. Then a + P is a unit in R/P and there is
b such that ab 1 P. It follows that the ideal (a) + P = R and therefore P is
maximal. The converse is similar.
1.3.4. Corollary. (1) A maximal ideal is a prime ideal.
(2) A ring is an domain if and only if the zero ideal is a prime ideal.
(3) A ring a eld if and only if the zero ideal is a maximal ideal.
1.3.5. Corollary. (1) If : R S is a ring homomorphism and Q S a prime
ideal then
1
(Q) is a prime ideal of R.
(2) Let I R be an ideal. An ideal I P is a prime ideal in R if and only if
P/I is a prime ideal in R/I.
(3) Let I R be an ideal. An ideal I P is a maximal ideal in R if and only if
P/I is a maximal ideal in R/I.
Proof. (1) By 1.2.9 R/
1
(Q) is a subring of the domain S/Q. (2) (3) By 1.2.11
R/P (R/I)/(P/I).
14 1. A DICTIONARY ON RINGS AND IDEALS
1.3.6. Example. An ideal in Z is a prime ideal if it is generated by 0 or a prime
number. Any nonzero prime ideal is a maximal ideal.
1.3.7. Denition. For an ideal I in a ring R the radical is

I = a R[a
n
I for some n
a is nilpotent is a
n
= 0 for some positive integer n. A ring is reduced if the
nilradical

0 = 0, that is if 0 is the only nilpotent element.


1.3.8. Proposition. (1) The radical of an ideal is an ideal.
(2) The nilradical is contained in any prime ideal.
(3) A domain is reduced.
Proof. (1) If a
m
, b
n
I then by the binomial formula
(a +b)
m+n
=
m+n

k=0
_
m+n
k
_
a
m+nk
b
k
I
and the radical is an ideal. (2) (3) Clearly a nilpotent element is contained in any
prime ideal.
1.3.9. Exercise. (1) Show that the characteristic of a domain is either 0 or a prime
number.
(2) Let m, n be a natural numbers. Show that n + (m) Z/(m) is a unit if and only if
m, n are relative prime.
(3) Let m be a natural number. Show that Z/(m) is reduced if m is square free.
(4) Show that a product of reduced rings is reduced.
(5) Let a be nilpotent. Show that 1 a is a unit.
(6) Let I, J be ideals. Show that

IJ =

I J =

J.
(7) Assume an ideal I is contained in a prime ideal P. Show that

I P.
1.4. Chinese remainders
1.4.1. Denition. Ideals I, J R are comaximal ideals if I +J = R.
1.4.2. Theorem (Chinese remainder theorem). Let I
1
, . . . , I
k
be pairwise comax-
imal ideals in a ring R. Then
(1) For a
1
, . . . , a
k
Rthere is a a R, such that aa
m
I
m
for m = 1, . . . , k
(2)
I
1
I
k
= I
1
I
k
(3) The product of projections
R/I
1
I
k
R/I
1
R/I
k
is an isomorphism.
Proof. (1) For each m
R =

n=m
(I
m
+I
n
) = I
m
+

n=m
I
n
So choose u
m
I
m
and v
m

n=m
I
n
with u
m
+v
m
= 1. Put a = a
1
v
1
+ +
a
k
v
k
. Then aa
m
= +a
m
u
m
+ I
m
. (2) For a in the intersection assume
by induction that a I
2
I
k
. From the proof of (1) a = u
1
a + av
1
I
1
I
k
.
(3) Surjectivity follows from (1). The kernel is the intersection which by (2) is the
product. 1.2.9 gives the isomorphism.
1.5. UNIQUE FACTORIZATION 15
1.4.3. Corollary. Let P
1
, . . . , P
k
be pairwise different maximal ideals in a ring R.
Then
P
n
1
1
P
n
k
k
= P
n
1
1
P
n
k
k
and
R/P
n
1
1
P
n
k
k
R/P
n
1
1
R/P
n
k
k
is an isomorphism.
1.4.4. Denition. An element e in a ring R is idempotent if e = e
2
. A nontrivial
idempotent is an idempotent e ,= 0, 1.
1.4.5. Proposition. A ring R is a product of two nonzero rings if and only if it
contains a nontrivial idempotent e.
Proof. Use that the ideals Re and R(1 e) are proper and comaximal.
1.4.6. Exercise. (1) Show that for a prime number p the rings Z/(p
2
) and Z/(p)
Z/(p) are not isomorphic.
(2) Let n = p
n
1
1
. . . p
n
k
k
be a factorization into different primes. Show that
Z/(p
n
1
1
p
n
k
k
) Z/(p
n
1
1
) Z/(p
n
k
k
)
is an isomorphism.
(3) Let elements e
1
+ e
2
= 1 with e
1
e
2
= 0 be given in a ring R. Show that R
R/(e
1
) R/(e
2
).
(4) Let I, J be ideals such that

I,

J are comaximal. Show that I, J are comaximal.


1.5. Unique factorization
1.5.1. Lemma. Let R be a domain and (a) = (b) principal ideals. Then there is a
unit u R such that b = ua.
Proof. b = ua and a = vb giving b = uvb. If b ,= 0 then uv = 1 showing that u is
a unit.
1.5.2. Denition. Let R be a domain and T the set of principal ideals different
from (0) and R. By 1.5.1 multiplication of generators gives a well dened multi-
plication of principal ideals on T. Inclusion in T, (a) (b) is division and written
b[a. An element in T maximal for inclusion is an irreducible principal ideal. A
generator of an irreducible element is an irreducible element in R. A principal
prime ideal is a prime divisor. A generator of a prime divisor is a prime element
in R. Clearly a prime divisor is irreducible.
1.5.3. Denition. A domain R is a unique factorization domain if
(1) Every irreducible ideal in T is a prime divisor.
(2) Every element in T is a product of irreducible elements.
1.5.4. Proposition. In a unique factorization domain the factorization of elements
in T into prime divisors is unique up to order.
Proof. Proceed by induction on the shortest factorization of an element in T. Let
(a
1
) . . . (a
m
) = (b
1
) . . . (b
n
) be factorizations into irreducibles. (a
1
) is a prime
ideal, so by 1.3.2 and reordering (b
1
) = (a
1
). By cancellation m1 = n 1 and
(a
i
) = (b
i
) after a reordering.
1.5.5. Denition. A domain R is a principal ideal domain if every ideal is princi-
pal.
16 1. A DICTIONARY ON RINGS AND IDEALS
1.5.6. Theorem. A principal ideal domain R is a unique factorization domain.
Proof. Let (a) be irreducible and x, y / (a). Then (a, x), (a, y) are principal ideals
properly containing (a) giving (a, x) = (a, y) = R. Let ba + cx = da + ey = 1
and look at (ba+cx)(da+ey) = 1 to see that xy / (a). It follows that (a) is prime,
part (1) of 1.5.3. If (b) is not irreducible, then (b) = (a
1
)(b
1
) for some irreducible
(a
1
) and (b) (b
1
). Continue this process to get (b) = (a
1
) . . . (a
n
)(b
n
) for some
irreducibles (a
1
), . . . (a
n
). The chain of ideals (b) (b
1
) (b
n
) has a
union

n
(b
n
) which is a principal ideal. The generator of the union must be in
some (b
n
). Therefore (b
n
) = (b
n+1
) giving that (b
n
) is irreducible. This gives a
factorization required in 1.5.3 part (2).
1.5.7. Example. The integers Z is a principal ideal domain and therefore a unique
factorization domain.
1.5.8. Denition. The supremum in the set of all principal ideals of a set of ele-
ments in T is the greatest common divisor and an inmum is the least common
multiple.
1.5.9. Corollary. In a unique factorization domain the greatest common divisor
and the least common multiple of a nite set of elements exist.
If (a) = (p
1
)
m
1
. . . (p
k
)
m
k
and (b) = (p
1
)
n
1
. . . (p
k
)
n
k
with m
i
, n
i
0 then great-
est common divisor is (p
1
)
min(m
1
,n
1
)
. . . (p
k
)
min(m
k
,n
k
)
and least common multiple
is (p
1
)
max(m
1
,n
1
)
. . . (p
k
)
max(m
k
,n
k
)
.
1.5.10. Exercise. (1) Showthat an irreducible element in a principal ideal domain gen-
erates a maximal ideal.
(2) Show that there are innitely many prime numbers.
(3) Let Z[

1] be the smallest subring of C containing

1. Show that Z[

1] is a
principal ideal domain.
(4) Let Z[

5] be the smallest subring of Ccontaining

5. Show that Z[

5] is not
a unique factorization domain.
1.6. Polynomials
1.6.1. Denition. Let R be a ring. The polynomial ring R[X] is the additive
group given by the direct sum

n
RX
n
, n = 0, 1, 2, . . . consisting of all nite
sums f = a
0
+ a
1
X + . . . a
m
X
m
, a polynomial with a
n
R being the n

th
coefcient. Multiplication is given by X
i
X
j
= X
i+j
extended by linearity. If
g = b
0
+b
1
X +. . . b
n
X
n
is an other polynomial, then
f +g = (a
0
+b
0
) + (a
1
+b
1
)X + + (a
k
+b
k
)X
k
+. . .
fg = a
0
b
0
+ (a
0
b
1
+a
1
b
0
)X + + (a
0
b
k
+a
1
b
k1
+ +a
k
b
0
)X
k
+. . .
A monomial is polynomial of form aX
n
. The construction may be repeated to
give the polynomial ring in n-variables R[X
1
, . . . , X
n
] or even in innitely many
variables.
1.6.2. Denition. The degree, deg(f), of a polynomial 0 ,= f R[X] is the
index of the highest nonzero coefcient, the leading coefcient. A polynomial
with leading coefcient the identity is a monic polynomial.
1.6.3. Remark. (1) R is identied with the subring of constants in the polyno-
mial ring R[X
1
, . . . , X
n
].
(2) The nonzero constants are the polynomials of degree 0.
1.6. POLYNOMIALS 17
(3) The constant polynomial 1 is the unique monic polynomial of degree 0 and
the identity in the polynomial ring.
1.6.4. Proposition. Let 0 ,= f, g R[X].
(1) If fg ,= 0 then deg(fg) deg(f) + deg(g).
(2) If the leading coefcient of f or g is a nonzero divisor in R, then fg ,= 0 and
deg(fg) = deg(f) + deg(g)
Proof. (1) This is clear. (2)Clearly the leading coefcient of the product is the
product of the leading coefcients.
1.6.5. Corollary. Let R be a domain.
(1) The polynomial ring R[X] is a domain.
(2) The units in R[X] are the constants, which are units in R.
1.6.6. Theorem. Let R be a domain, 0 ,= f, d R[X] polynomials with d monic.
Then there are a unique q, r R[X] such that
f = qd +r, r = 0 or deg(r) < deg(d)
Proof. Induction on deg(f). If deg(f) < deg(d) then q = 0, r = f. Otherwise
if a is the leading coefcient of f, then f adX
deg(f)deg(d)
has degree less than
deg(f). By induction f adX
deg(f)deg(d)
= qd +r giving the claim.
1.6.7. Proposition. Let : R S be a ring homomorphism. For any element
b S there is a unique ring homomorphism R[X] S extending and mapping
X b.
Proof.
a
0
+a
1
X +. . . a
m
X
m
(a
0
) +(a
1
)b +. . . (a
m
)b
m
is clearly the one and only choice.
1.6.8. Denition. The homomorphism in 1.6.7 is the evaluation map at b in S.
The image of a polynomial f R[X] is denoted f(b) S.
1.6.9. Proposition. Let I Rbe an ideal. Then there is a canonical isomorphism.
(R/I)[X] R[X]/IR[X]
Proof. There is an obvious pair of inverse homomorphisms constructed by 1.2.9
and 1.6.7.
1.6.10. Corollary. If P R is a prime ideal, then PR[X] R[X] is a prime
ideal.
1.6.11. Denition. Let : R S be a ring homomorphism and B S a subset.
The ring generated over R by B is
R[B] = (R)[B] S
the smallest subring of S containing (R) B. If there is a nite subset B such
that R[B] = S then S is a nite type ring or a nitely generated ring over R.
1.6.12. Corollary. Let : R S be a ring homomorphism.
18 1. A DICTIONARY ON RINGS AND IDEALS
(1) If b

S and X

is a family of variables, then there is a surjective ring


homomorphism
R[X

] R[b

], X

making R[b

] a factor ring of the polynomial ring R[X

].
(2) If S is a nite type ring over R, then S is a factor ring of a polynomial ring
in nitely many variables over R.
1.6.13. Exercise. (1) Let K be a eld. Show that there are innitely many prime ideals
in K[X].
(2) What are the units in the ring Z[X]/(1 2X)?
(3) Determine the prime ideals in Q[X]/(X X
2
).
(4) Show that the ring Z[X] is not a principal ideal domain.
(5) Show that the ring Q[X, Y ] is not a principal ideal domain.
1.7. Roots
1.7.1. Denition. Let : R S be a ring homomorphism and f R[X] a
polynomial. An element b S is a root of f (in S) if the evaluation f(b) = 0.
1.7.2. Proposition. Let R be a domain. An element a R is a root of the polyno-
mial f R[X] if and only if there is a q R[X] such that
f = q(X a)
Proof. By 1.6.6 f = q(Xa)+r. It follows that a is a root if and only if r = 0.
1.7.3. Corollary. Let R is a domain. There are at most deg(f) roots in a nonzero
polynomial f R[X].
1.7.4. Denition. The multiplicity of a root a of a nonzero polynomial f R[X]
is highest m such that
f = q(X a)
m
A root of multiplicity m = 1 is a simple root.
1.7.5. Corollary. Let R is a domain. If m
1
, . . . , m
k
are the multiplicities of the
roots of a nonzero polynomial f R[X], then m
1
+ +m
k
deg(f).
1.7.6. Denition. The derivative of a polynomial f =

a
n
X
n
R[X] is
f

na
n
X
n1
1.7.7. Lemma. The derivative satises
(1) (f +g)

= f

+g

.
(2) (fg)

= f

g +fg

(3) If f is constant, then f

= 0.
1.7.8. Proposition. Let R is a domain. An element a R is a root of multiplicity
m > 1 of a nonzero f R[X] if and only if a is a root of f and f

.
Proof. By 1.6.6 f = q(Xa)
2
+cX+d and by 1.7.7 f

= q

(Xa)
2
+2q(X
a) +c. I follows that a is a root of multiplicity m > 1 if and only if c = d = 0.
1.7.9. Exercise. (1) Let a
1
, . . . a
k
be roots with multiplicities m
1
, . . . , m
k
in a poly-
nomial f. Show that m
1
+ +m
k
deg(f).
(2) Let K be a eld and let a
1
, . . . , a
n
K. Show that the ideal (X
1
a
1
, . . . , X
n
a
n
)
is maximal in K[X
1
, . . . , X
n
].
(3) Let the characteristic char(R) = n > 0. What is (X
n
)

in R[X].
1.8. FIELDS 19
1.8. Fields
1.8.1. Denition. Let p Z be a prime number. The factor ring F
p
= Z/(p) is a
eld with p elements. Together with Qthey constitute the prime elds.
1.8.2. Theorem. Let K be a eld then the polynomial ring K[X] is a principal
ideal domain.
Proof. Let d ,= 0 be a polynomial of lowest degree in an ideal I. Given f I then
by 1.6.6 f = qd + r with r = f qd I. By degree considerations r = 0 and
I = (d).
1.8.3. Corollary. Let K be a eld then the polynomial ring K[X] is a unique
factorization domain.
Proof. Follows from 1.5.6.
1.8.4. Denition. A subeld is a subring, which is a eld. A eld extension is the
inclusion of a subeld K L in a eld. A nite eld extension K L is an
extension, where L is nite dimensional as a vector space over K.
1.8.5. Example. (1) Let K be a eld and f an irreducible polynomial in K[X].
Then K K[X]/(f) is a nite eld extension.
(2) Let K R L be a subring in a nite eld extension. Then R is a eld.
Namely multiplication on R with a nonzero element of R is a K-linear map
on the nite dimensional K-vector space R and therefore an isomorphism.
1.8.6. Proposition. (1) Let K be a eld and f a polynomial in K[X]. Then there
is a nite eld extension K L such that f factors in linear factors in L[X].
(2) If K L
1
and K L
2
are nite eld extensions then there is a nite eld
extension K L such that L
1
L
2
L.
Proof. (1) Assume f irreducible. In L = K[X]/(f) the class X + (f) is a root of
f. In general proceed adjoining roots of irreducible factors of f. (2) An element x
in a nite eld extension K K

is the root of the irreducible monic polynomial


f generating the kernel of the evaluation homomorphism K[X] K

, X x.
Now proceed by (1) adjoining elements in L
2
to L
1
.
1.8.7. Proposition. Let p Z be a prime number. For any power q = p
n
there is
a eld F
q
with q elements, unique up to isomorphism.
Proof. Let F
p
K be a eld extension, where X
q
X factors into linear fac-
tors, 1.8.6 (1). The subset of roots is the set of elements xed under n-times the
Frobenius and therefore a subring being a subeld by 1.8.5 (2). The derivative
(X
q
X)

= 1 so by 1.7.8 there are q elements in this subeld. Uniqueness


follows from 1.8.6 (2).
1.8.8. Exercise. (1) Show that the ring R[X]/(X
2
+ 1) is isomorphic to the eld of
complex numbers.
(2) Show that the ring F
2
[X]/(X
2
+X + 1) is a eld with 4 elements.
(3) Show that the ring F
2
[X]/(X
3
+X + 1) is a eld with 8 elements.
(4) Let K L be a eld extension of elds of characteristic 0 and let a L be a root of
an irreducible polynomial f K[X]. Show that a is a simple root.
(5) Let p be a prime number. Show that F
p
is the only ring with p elements.
(6) Let p be a prime number. Show that a ring with p
2
elements is isomorphic to one of
four non isomorphic Z/(p
2
), F
p
F
p
, F
p
[X]/(X
2
), F
p
2.
20 1. A DICTIONARY ON RINGS AND IDEALS
1.9. Power series
1.9.1. Denition. Let R be a ring. The power series ring R[[X]] is the additive
group

n
RX
n
, n = 0, 1, 2, . . . of all power series

a
n
X
n
with n

th coefcient
a
n
R. Multiplication is given by X
i
X
j
= X
i+j
extended by linearity. For
another power series

b
n
X
n
the sum and product are

a
n
X
n
+

b
n
X
n
=

(a
n
+b
n
)X
n

a
n
X
n

b
n
X
n
=

k
a
nk
b
k
)X
n
The construction may be repeated to give the power series ring in n-variables
R[[X
1
, . . . , X
n
]] or even in innitely many variables.
1.9.2. Remark. The polynomial ring is identied as a subring R[X] R[[X]] of
power series with only nitely many nonzero terms.
1.9.3. Denition. The order, o(f), of a power series 0 ,= f R[[X]] is the index
of the least nonzero coefcient.
1.9.4. Proposition. If R is a domain, then R[[X]] is a domain and for 0 ,= f, g
R[X]
o(fg) = o(f) +o(g)
Proof. Clearly the lowest nonzero coefcient in the product is the product of the
two lowest nonzero coefcients.
1.9.5. Proposition. A power series f =

a
n
X
n
is a unit if and only if a
0
is a
unit.
Proof. It sufces to look at a power series f = 1 gX. Then the power series
1/f = 1 +gX +g
2
X
2
+ +g
n
X
n
+. . . is well dened and f 1/f = 1.
1.9.6. Theorem. If K is a eld, then K[[X]] is a principal ideal domain. and (X)
is the only nonzero prime ideal.
Proof. If the lowest order of an element in an ideal I is n. Then clearly I =
(X
n
).
1.9.7. Corollary. If K is a eld, then K[[X]] is a unique factorization domain.
1.9.8. Proposition. Let I R be an ideal. Then there is a canonical surjective
homomorphism
R[[X]]/IR[[X]] R/I[[X]]
1.9.9. Corollary. If Q R[[X]] is a maximal ideal, then P = Q R R is a
maximal ideal and Q = (P, X).
Proof. X Q so R/Q R R[[X]]/Q.
1.9.10. Exercise. (1) Show that the ring Z[[X]] is not a principal ideal domain.
(2) Show that the ring Q[[X, Y ]] is not a principal ideal domain.
(3) Let K be a eld. Show that (X
1
, . . . , X
n
) is the unique maximal ideal in the power
series ring K[[X
1
, . . . , X
n
]].
(4) Let a R be nilpotent. Show that the ring R[[X]]/(X a) is isomorphic to R.
(5) What is R[[X]]/(X a) if a R is a unit?
(6) Let I R be an ideal. Show that IR[[X]] R[[X]] is not a maximal ideal.
2
Modules
2.1. Modules and homomorphisms
2.1.1. Denition. Let R be a ring. A module (R-module) is an abelian group
M, addition (x, y) x + y and zero 0, together with a scalar multiplication
R M M, (a, x) ax satisfying
(1) associative : (ab)x = a(bx)
(2) bilinear : a(x +y) = ax +ay, (a +b)x = ax +bx
(3) identity: 1x = x
for all a, b R, x, y M. A submodule M

M is an additive subgroup such


that ax M

for all a R, x M

. A homomorphism is an additive group


homomorphism f : M N respecting scalar multiplication
f(x +y) = f(x) +f(y), f(ax) = af(y)
for all a R, x, y M. An isomorphism is a homomorphism f : M N
having an inverse map f
1
: N M which is also a homomorphism. The
identity isomorphism is denoted 1
M
: M M.
2.1.2. Lemma. Let R be a ring and M a module.
(1) a0 = 0 = 0x
(2) (1)x = x
(3) (a)x = (ax) = a(x)
for all a R, x M.
Proof. (1) Calculate a0 = a(0 +0) = a0 +a0 and cancel to get a0 = 0. Similarly
0x = 0. (2) By (1) 0 = 0x = (1 + (1))x = x + (1)x, so conclude x =
(1)x. (3) Calculate (a)x = ((1)a)x = (1)(ax) = (ax). Similarly
a(x) = (ax).
2.1.3. Lemma. Let R be a ring and f : M N a homomorphism of modules.
(1) f(0) = 0.
(2) f(ax +by) = af(x) +bf(y).
(3) f(x) = f(x) for all x M.
for all a, b R, x, y M.
Proof. (1) Calculate f(0) = f(0 + 0) = f(0) + f(0) and conclude f(0) = 0.
(2) Calculate f(ax + by) = f(ax) + f(by) = af(x) + bf(y). (3) By 2.1.2
f(x) = f((1)x) = (1)f(x) = f(x).
2.1.4. Example. (1) The zero group is the zero module.
(2) Over the zero ring the zero module is the only module.
(3) The zero subgroup of a module is the zero submodule.
(4) The ring R is a module under multiplication. An ideal is a submodule.
21
22 2. MODULES
(5) If R is a eld, a module is a vector space and a homomorphism is a linear
map.
(6) A module over Z is an abelian group and an additive map of abelian groups
is a homomorphism.
2.1.5. Proposition. A bijective homomorphism is an isomorphism.
Proof. Let f : M N be a bijective homomorphism of R-modules and let g :
N M be the inverse map. For x, y N write x = f(g(x)), y = f(g(y)) and
get additivity of g, g(x + y) = g(f(g(x)) + f(g(y))) = g(f(g(x) + g(y))) =
g(x) +g(y). Similarly for a R g(ax) = g(af(g(x))) = g(f(ag(x))) = ag(x),
so g respects scalar multiplication and is a homomorphism.
2.1.6. Lemma. Let a R and M be a module. The map M M, x ax is a
homomorphism.
Proof. Let f(x) = ax and calculate f(x+y) = a(x+y) = ax+ay = f(x)+f(y)
and f(bx) = a(bx) = (ab)x = (ba)x = b(ax) = bf(x) to get that f is a
homomorphism. Remark that the last calculation uses that R is commutative 1.1.2
(4).
2.1.7. Denition. Let a R and M be a module.
(1) The scalar multiplication with a is the homomorphism, 2.1.6,
a
M
: M M, x ax
(2) a R is a nonzero divisor on M if scalar multiplication a
M
is injective, i.e.
ax ,= 0 for all 0 ,= x M. Otherwise a is a zero divisor.
2.1.8. Remark. The two notions of nonzero divisor on R : 1.1.6 as element in the
ring and 2.1.4 as scalar multiplication on the ring coincide.
2.1.9. Example. If Ris a eld, then scalar multiplication on a vector space is either
zero or an isomorphism.
2.1.10. Lemma. Let : R S be a ring homomorphism and N an S-module.
The map
R N N, (a, x) (a)x
is an R-scalar multiplication on N, 2.1.1.
Proof. Let a, b R, x, y N and (a, x) = (a)x. Calculate (a + b, x) =
(a + b)x = (a)x + (b)x = (a, x) + (b, x), (a, x + y) = (a)(x + y) =
(a)x + (a)y = (a, x) + (a, y), (1, x) = (1)x = 1x = x and (ab, x) =
(ab)x = (a)(b)x = (a, (bx)) showing the conditions 2.1.1.
2.1.11. Denition. Let : R S be a ring homomorphism. The restriction of
scalars of an S-module N is the same additive group N viewed as an R-module
through . The scalar multiplication is 2.1.10
R N N, (a, x) ax = (a)x
An S-module homomorphism g : N N

is also an R-module homomorphism.


2.1.12. Example. (1) The scalar multiplication with a Restriction of scalars for
the unique ring homomorphismZ Rgive just the underlying abelian group
of a module, 2.1.4 (6).
2.2. SUBMODULES AND FACTOR MODULES 23
(2) Let I R be an ideal. Restriction of scalars along the projection R R/I
gives any R/I-module as an R-module.
2.1.13. Proposition. Let R be a ring. There is a dictionary:
(1) To an R[X]-module N associate the pair (N, f) consisting of N as R-module
through restriction of scalars and f = X
N
: N N, f(y) = Xy scalar
multiplication with X as an R-module homomorphism. An R[X]-homomorphism
g : N N

gives an R-homomorphism such that g f = f

g.
(2) To a pair (N, f) of an R-module and a homomorphismf : N N associate
the R[X]-module with abelian group N and scalar multiplication determined
by Xy = f(y) for y N. Note
(

a
n
X
n
)y =

a
n
f
n
(y)
An R-homomorphism g : N N

such that g f = f

g is an R[X]-
homomorphism.
Proof. The statement is an algorithm to follow.
2.1.14. Proposition. Let R be a ring and M a module. The abelian group RM
with multiplication
(a +x)(b +y) = ab + (ay +bx)
is a ring. R is a subring and M is an ideal.
Proof. Simple calculations show that the conditions 1.1.2 are satised.
2.1.15. Exercise. (1) Showthat a composition of homomorphisms is a homomorphism.
(2) Show that composition of scalar multiplications with a, b R on a module M is a
scalar multiplication with the product, a
M
b
M
= (ab)
M
.
(3) Let : R S be a ring homomorphism. Show that is an R-module homomor-
phism, when S is viewed as R-module through restriction of scalars 2.1.11.
(4) Fill out the dictionary 2.1.13.
2.2. Submodules and factor modules
2.2.1. Lemma. Let R be a ring and M a module. Let N

be a family of submod-
ules. Then the additive subgroups

and

are submodules.
Proof. Use the formulas

(x

+ y

) and a

ax

to
conclude that

is a submodule. If x, y N

for all , then x + y, ax N

for all , so

is a submodule.
2.2.2. Denition. Let Rbe a ring and M a module. The intersection of all submod-
ules containing a subset Y M is the submodule generated by Y and denoted
RY . This is the smallest submodule, 2.2.1, of M containing Y . The module M is
generated by Y if RY = M. Let I be an ideal. The submodule generated by all
products ax, a I, x M is denoted IM.
2.2.3. Proposition. Let R be a ring and M a module. If Y M, then RY =

yY
Ry,
RY = a
1
y
1
+ +a
n
y
n
[a
i
R, y
i
Y
Proof. The righthand side is contained in the submodule RY . Moreover the right-
hand side is a submodule containing Y , so equality.
2.2.4. Corollary. Let I R be an ideal and M a module.
24 2. MODULES
(1)
IM = a
1
y
1
+ +a
n
y
n
[a
i
I, y
i
M
(2) If I = (a) is principal, then
aM = (a)M = ay[y M
Proof. (1) is clear. (2) By (1) an element in aM is

b
i
ay
i
= a

b
i
y
i
= ay for
y =

b
i
y
i
M as claimed.
2.2.5. Lemma. Let Rbe a ring, M a module and N M a submodule. Let M/N
be the abelian factor group, then the map
R M/N M/N, (a, x +N) ax +N
is well dened and a scalar multiplication, 2.1.1.
Proof. If x+N = y+N then xy N and so a(xy) = axay N. Therefore
ax + N = ay + N and the multiplication is well dened. Since representatives
may be chosen such that (x +N) +(y +N) = x +y +N, a(x +N) = ax +N,
the laws for scalar multiplication are satised.
2.2.6. Denition. Let Rbe a ring, M a module and N M a submodule, then the
factor module is the additive factor group M/N with, 2.2.5, scalar multiplication
a(x +N) = ax +N. The projection p : M M/N, x x +N is a surjective
homomorphism.
2.2.7. Lemma. Let R be a ring, N M a submodule and p : M M/N
the projection. p is surjective and if Y M generates M, then p(Y ) M/N
generates the factor module.
Proof. Clearly if RY = M then Rp(Y ) = p(RY ) = M/N.
2.2.8. Example. (1) A submodule of R is the same as an ideal.
(2) Both an ideal I R and a factor ring R/I are modules.
(3) The module structure on R/I as a factor module and the structure by restric-
tion of scalars through the projection R R/I are identical.
2.2.9. Proposition. Let R = R
1
R
2
be the product ring 1.1.4. There is a bijective
(up to natural isomorphism) correspondence.
(1) If M
1
is an R
1
-module and M
2
is an R
2
-module, then M = M
1
M
2
is an
R-module with coordinate scalar multiplication. A pair of homomorphisms
induce a homomorphism on the product.
(2) If M is an R-module then M
1
= (1, 0)M is an R
1
-module and M
2
=
(0, 1)M is an R
2
-module. A homomorphism induces a pair of homomor-
phisms.
2.2.10. Remark. The correspondence 2.2.9 indicates that the structure of modules
and homomorphisms over a product ring is identied with the structure of pairs of
modules and homomorphisms over each component ring in the product.
2.2.11. Exercise. (1) Give an example of two submodules N, L M such that the
union N L is not a submodule.
(2) Let R be a ring and a R. Show that the R-module R[X]/(X a) is isomorphic
to R.
(3) Show that the projection p 2.2.6 is a homomorphism.
(4) Fill in the details in the dictionary 2.2.9
2.3. KERNEL AND COKERNEL 25
2.3. Kernel and cokernel
2.3.1. Lemma. Let R be a ring and f : M N a homomorphism of modules.
Given submodules M

M, N

N, then f
1
(N

) M and f(M

) N are
submodules.
Proof. If x, y f
1
(N

) then f(x + y) = f(x) + f(y) N

and for a R
f(ax) = af(x) N

so x+y, ax f
1
(N

) proving f
1
(N

) to be a submodule.
The same equations prove that f(M

) is a submodule.
2.3.2. Denition. Let f : M N be a homomorphism of modules. Then there
are submodules, 2.3.1.
(1) The kernel Ker f = f
1
(0).
(2) The image Imf = f(M).
(3) The cokernel Cok f = N/ Imf.
2.3.3. Proposition. Let f : M N be a homomorphism of modules.
(1) f is injective if and only if Ker f = 0.
(2) f is surjective if and only if Cok f = 0.
(3) f is an isomorphism if and only if Ker f = 0 and Cok f = 0.
Proof. (1) If f is injective and x Ker f then f(x) = 0 = f(0) so x = 0.
Conversely if Ker f = 0 and f(x) = f(y) then f(x y) = 0 so x = y. (2) The
factor module N/ Imf = 0 if and only if Imf = N. (3) This follows from (1)
and (2).
2.3.4. Example. Let a R give scalar multiplication a
M
: M M, x ax.
(1) Ima
M
= aM = ax M[x M.
(2) Ker a
M
= x M[ax = 0.
(3) Cok a
M
= M/aM.
2.3.5. Theorem. Let f : M N be a homomorphism of modules.
(1) Let L Ker f be a submodule. Then there is a unique homomorphism f

:
M/L N such that f = f

p.
M
p

N
M/L
f

(2) The homomorphism f

: M/ Ker f N is a module isomorphism onto the


submodule Imf of N.
M
p

Imf
M/ Ker f
f

Proof. (1) If x+L = x

+L then xx

L giving f(x) = f(x

). The factor map


f

: M/L N, x + L f(x) is therefore well dened and f = f

p. Since
f, p are homomorphisms and p is surjective it follows that f

is a homomorphism.
(2) The kernel of f

is Ker f/ Ker f = 0 so by 2.3.3 it is an isomorphism.


26 2. MODULES
2.3.6. Corollary. Let p : M M/N be the projection onto a factor module. The
map L

L = p
1
(L

) gives a bijective correspondence between submodules


in M/N and submodules in M containing N. Also L

= p(L) = L/N. This


correspondence preserves inclusions, additions and intersections of submodules.
Proof. If L

is a submodule of M/N then clearly p(p


1
(L

)) = L

. If N L
is a submodule of M then clearly L p
1
(p(L)). Moreover if x p
1
(p(L))
then p(x) = p(y) for some y L and therefore y x N. It follows that
L = p
1
(p(L)) and the correspondence is bijective. Inclusions are easily seen to
be preserved. Also easily p(L
1
+ L
2
) = p(L
1
) + p(L
2
) and p
1
(L

1
L

2
) =
p
1
(L

1
) p
1
(L

2
) hold. The two resting equalities are consequences of this and
bijectivity of the correspondence.
2.3.7. Corollary. Let L N M be submodules. Then there is a canonical
isomorphism
M/N (M/L)/(N/L)
Proof. The kernel of the surjective east-south composite
M

M/L

M/N

(M/L)/(N/L)
is N. By 2.3.5 the horizontal lower factor map gives the isomorphism.
2.3.8. Corollary. Let L, N M be submodules. Then there is a canonical iso-
morphism
N/N L N +L/L
given by x +N L x +L.
Proof. The kernel of the east-south composite
N

N +L

N/N L

N +L/L
is N L. Since x + y + L = x + L for x N, y L this composite is also
surjective. By 2.3.5 the horizontal lower factor map gives the isomorphism.
2.3.9. Proposition. Let f : M N and g : N L be homomorphisms such that
Imf Ker g. Then there is a unique homomorphism g

: Cok f L such that


g = g

p.
M
f

N
p

L
Cok f
g

Proof. This follows from 2.3.5.


2.3.10. Lemma. Let R be a ring and M a module. For x M the map R
M, a ax is the unique homomorphism such that 1 x.
2.3. KERNEL AND COKERNEL 27
Proof. Let f(a) = ax. f(ab) = (ab)x = a(bx) = af(b) shows that f is a
homomorphism. This argument does not use the commutativity 1.1.2 of R.
2.3.11. Denition. Denote the homomorphism 2.3.10
1
x
: R M, a ax
The annihilator of x is the ideal
Ann(x) = Ker 1
x
= a R[ax = 0
For a subset Y M the annihilator Ann(Y ) =

yY
Ann(y) is the ideal of
elements
Ann(Y ) = a R[ay = 0, for all y Y
2.3.12. Proposition. Let R be a ring and Y a subset of a module M.
(1) Ann(Y ) = Ann(RY ).
(2) a Ann(M) if and only if a
M
= 0.
(3) If modules M M

then Ann(M) = Ann(M

).
(4) The induced homomorphism
1

x
: R/ Ann(x) Rx, a + Ann(x) ax
is an isomorphism.
Proof. (1) If a Ann(Y ) then a

b
i
y
i
=

b
i
ay
i
= 0 giving the not so obvious
Ann(Y ) Ann(RY ). (2) Clear since a
M
(x) = ax, 2.1.7. (3) Let f : M M

be an isomorphism and a R. Then af(x) = f(ax) expresses that f a


M
=
a
M
f. Since f is bijective, a
M
= 0 if and only if a
M
= 0. By (2) Ann(M) =
Ann(M

). (4) This follows from 2.3.5.


2.3.13. Corollary. Let I, J R be ideals.
(1) Ann(R/I) = I
(2) If R/I R/J then I = J.
2.3.14. Lemma. Let I R be an ideal and M an R-module. If I Ann(M)
then M is an R/I-module with the scalar multiplication
R/I M M, (a +I, x) ax
That is, M is an R/ Ann(M)-module.
2.3.15. Example. Let a R give scalar multiplication a
M
: M M, x ax.
(1) a Ann(Ker a
M
).
(2) a Ann(Cok a
M
).
(3) Ker a
M
and Cok a
M
are modules over the factor ring R/(a), 2.3.14.
2.3.16. Denition. Let R be a ring and L, N M submodules. The colon ideal
N : L is the ideal of elements a R such that aL N.
2.3.17. Proposition. The colon ideal of submodules L, N M is
N : L = Ann(N +L/N) = Ann(N/N L)
Proof. If a N : L then aL N. Therefore a(N +L) N and a Ann(N +
L/N). Conversely if a Ann(N +L/N) then aL N and therefore a N : L.
The last equality follows from 2.3.8.
2.3.18. Exercise. (1) Give an example of a homomorphism f : M N submodules
M
1
, M
2
M such that f(M
1
M
2
) ,= f(M
1
) f(M
2
).
28 2. MODULES
(2) Give an example of a homomorphism f : M N submodules N
1
, N
2
N such
that f
1
(N
1
+N
2
) ,= f
1
(M
1
) +f
1
(M
2
).
(3) Let R be a ring and M a module. Show that M may be regarded as an R/ Ann(M)-
module in a natural way.
(4) Let L, N M be submodules. Show that Ann(L +N) = Ann(L) Ann(N).
(5) Let f : M N be a surjective homomorphism. Show that Ann(M) Ann(N).
(6) Let f : M N be an injective homomorphism. Show that Ann(N) Ann(M).
2.4. Sum and product
2.4.1. Lemma. Let R be a ring and (M

) a family of modules. The product

is the abelian group of all families (x

), x

with term wise ad-


dition. The setting
r(x

) = (rx

)
is a scalar multiplication on

. The direct sum


is the subgroup
of

consisting of families with only nitely many nonzero terms. This is a


submodule.
Proof. The laws in 2.1.1 are true for each factor and therefore trivially veried for
the product and sum.
2.4.2. Denition. Let R be a ring and M

a family of modules. By 2.4.1 there are


modules and homomorphisms
(1) The direct product is

.
(2) The projections p

are the homomorphisms p

((x

)) =
x

.
(3) The direct sum is

. Elements in

are written as nite sums

.
(4) The injections i

: M

are the homomorphisms given by i

(x) =
(x

), where x

= x and x

= 0, ,= .
2.4.3. Theorem. Let R be a ring and M

a family of modules.
(1) Given a family of homomorphisms f

: L M

, then there exists a unique


homomorphism f : L

such that f

= p

f.
L
f

(2) Given a family of homomorphisms g

: M

L, then there exists a unique


homomorphism g :

L such that g

= g i

L
M

Proof. (1) f(y) = (f

(y)) is the unique homomorphism. (2) g(

) =

(x

)
is well dened since only nitely many x

,= 0 and a homomorphism.
2.4. SUM AND PRODUCT 29
2.4.4. Denition. A family of submodules M

M constitutes a direct sum if


any element x

has a unique nite representation


x =

, x

2.4.5. Proposition. The following conditions are equivalent:


(1) The family M

M constitutes a direct sum.


(2) The natural homomorphism

is an isomorphism.
(3) For all
M

=
M

= 0
Proof. (1) (2): This is clear. (1) (3): If x

=
x

=
M

,
then x

=
x

= 0. Therefore by uniqueness x

= 0. (3) (1): If

=
0, then x

=
x

=
M

= 0. for any . This shows


uniqueness.
2.4.6. Denition. Let R be a ring. A module isomorphic to a direct sum

R of
copies of the ring R is a free module.
A basis of a module, is a subset Y such that any element admits a unique nite
representation

, where a

R, y

Y .
The standard basis of

R consists of the elements e

, where each is a family


with 0 for ,= and exactly 1 at index .
2.4.7. Proposition. A module is free if and only if it admits a basis. If y

is a basis
of F then F = Ry

as direct sum and there is an isomorphism


f :

R F
where f(

) =

.
Proof. Given a free module f :

R F then y

= f(e

) is a basis. Con-
versely given a basis y

F then 1
y

: R F 2.3.11 is a family of homomor-


phisms giving a homomorphismf :

R F by 2.4.3. As f(

) =

it follows that f is bijective and therefore by 2.1.5 an isomorphism.


2.4.8. Remark. The polynomial ring R[X
1
, . . . , X
n
] is free as R-module.
2.4.9. Example. (1) A nonzero ideal is a free module if and only if it has a basis
consisting of a nonzero divisor. Namely if x
1
,= x
2
where in a basis then the
product x
1
x
2
has two different representations.
(2) Let I R be an ideal. The module R/I is free if and only if I = 0 or I = R.
2.4.10. Proposition. Any module over a eld is free. Conversely if any module
over a nonzero ring is free, then the ring is a eld
Proof. By Zorns lemma any vector space admits a basis. If I R is an ideal and
R/I is free, then I = Ann(R/I) is either 0 or R. So R is a eld.
30 2. MODULES
2.4.11. Proposition. Let F be a free module with basis y

. For a module M and


a family of elements x

M there is a unique homomorphism g : F M such


that g(y

) = x

given by g(

) =

.
Proof. The basis y

F gives the isomorphism 2.4.7 f :


R F. The
family 1
x

: R M gives a homomorphism g

R M by 2.4.3. Then
g = g

f
1
.
2.4.12. Corollary. Let M be an R-module and

M
R the free module with basis
e
x
indexed by x M. The homomorphism

M
R M,

a
x
e
x

a
x
x
is surjective identifying M as a factor module of a free module in a natural way.
2.4.13. Denition. A module is indecomposable if it is not isomorphic to a direct
sum of two nonzero submodules, otherwise decomposable.
2.4.14. Example. Q is an indecomposable Z-module. Namely if
m
1
n
1
,
m
2
n
2
are
nonzero numbers in two submodules, then n
1
m
2
m
1
n
1
= n
2
m
1
m
2
n
2
is a nonzero num-
ber in the intersection.
2.4.15. Exercise. (1) Show that if a ring is decomposable as a module, then it is the
product of two nonzero rings.
(2) Let M

be a nite family of modules. Show that

,
(3) Let N

be a family of submodules modules. Show that

/N

and that

M

/N

2.5. Homomorphism modules


2.5.1. Lemma. Let R be a ring and f, g : M N homomorphisms.
(1) (f +g)(x) = f(x) +g(x) is a homomorphism.
(2) If a R, then (af)(x) = af(x) is a homomorphism.
Proof. Calculate according to 2.1.1. (1) (f +g)(x +y) = f(x +y) +g(x +y) =
f(x) + f(y) + g(x) + g(y) = (f + g)(x) + (f + g)(y) and (f + g)(ax) =
f(ax) +g(ax) = a(f(x) +g(x)) = a(f +g)(x). (2) (af)(x+y) = af(x+y) =
af(x) + af(y) = (af)(x) + (af)(y) and (af)(bx) = af(bx) = abf(x) =
baf(x) = b(af)(x). The last calculation uses that R is commutative 1.1.2 (4).
2.5.2. Denition. Let R be a ring and M, N modules. By 2.5.1, the homomor-
phism module Hom
R
(M, N) is the additive group of all homomorphism with
scalar multiplication
R Hom
R
(M, N) Hom
R
(M, N), (a, f) af = [x af(x)]
2.5.3. Denition. Let a R be a ring and f : M M

, g : N N

, h, k :
M

N homomorphisms of modules. By 2.5.1


(1) (h +k) f = h f +k f.
(2) (ah) f = a(h f).
(3) g (h +k) = g h +g k.
(4) g (ah) = a(g h).
2.5. HOMOMORPHISM MODULES 31
There is induced a homomorphism
Hom(f, g) : Hom
R
(M

, N) Hom
R
(M, N

)
(h : M

N) (g h f : M N

)
of R-modules.
2.5.4. Denition. Let R, S be rings. A functor is a construction T, which to an
R-module M associates S-modules T(M)
M T(M)
and for two R-modules M, N there is an additive homomorphism
Hom
R
(M, N) Hom
S
(T(M), T(N))
f T(f)
such that
(1) T(1
M
) = 1
T(M)
(2) T(g f) = T(g) T(f)
In case the homomorphism changes domain and target
Hom
R
(M, N) Hom
S
(T(N), T(M))
f T(f)
and
(1) T(1
M
) = 1
T(M)
(2) T(g f) = T(f) T(g)
the functor is contravariant.
Clearly (contravariant) functors transform isomorphisms into isomorphisms.
Given functors T, T

a natural homomorphism is a family


M
: T(M) T

(M)
of homomorphisms, such that for each f : M N the following diagram com-
mutes
T(M)
T(f)

(M)
T

(f)

T(N)

N

(N)
In the contravariant case the diagram is
T(M)

M

(M)
T(N)
T(f)

(N)
T

(f)

A natural isomorphism is a natural homomorphism such that each


M
is an iso-
morphism.
2.5.5. Proposition. Let R be a ring.
(1) The construction
N Hom
R
(M, N)
g Hom(1
M
, g)
32 2. MODULES
is a functor and
Hom
R
(N, N

) Hom
R
(Hom
R
(M, N), Hom
R
(M, N

))
is an R-homorphism.
(2) The construction
M Hom
R
(M, N)
f Hom(f, 1
N
)
is a contravariant functor and
Hom
R
(M, M

) Hom
R
(Hom
R
(M

, N), Hom
R
(M, N))
is an R-homorphism.
Proof. By 2.5.3 the construction on homomorphisms are homomorphisms. Given
also homomorphisms f

: M

and g

: N

. Then
Hom(1
M
, g

g) = Hom(1
M
, g

) Hom(1
M
, g)
and
Hom(f

f, 1
N
) = Hom(f, 1
N
) Hom(f

, 1
N
)
showing the conditions on compositions. Clearly also
Hom(1, g +g

) = Hom(1, g) + Hom(1, g

)
Hom(1, ag) = a Hom(1, g)
and similarly for the resting properties.
2.5.6. Corollary. Let a R give scalar multiplications a
M
, a
N
.
(1) The three constructions
Hom(a
M
, 1
N
) = Hom(1
M
, a
N
) = a
Hom
R
(M,N)
are the same homomorphism
Hom
R
(M, N) Hom
R
(M, N) f af
(2) The map R Hom
R
(M, M), a a
M
is a homomorphism.
2.5.7. Example. Let R = R
1
R
2
be the product ring. The constructions in 2.2.9
is: (1) A functor which to a pair of an R
1
-module and an R
2
-module associates an
R-module.(2) A functor which to an R-module associates a pair of an R
1
-module
and an R
2
-module.
2.5.8. Proposition. Let R be a ring and M

a family of modules. For any module


N there are natural isomorphisms
(1) Hom
R
(

, N)

Hom
R
(M

, N)
(2) Hom
R
(N,

Hom
R
(N, M

)
Proof. This is 2.4.3 reformulated. (1) A homomorphism g :

N is
uniquely determined by the family g

= g i

: M

N. (2) A homomorphism
f : N

is uniquely determined by the family f

= p

f : N M

.
2.5.9. Lemma. Let R be a ring and M, N modules. For x M there is a homo-
morphism Hom
R
(M, N) N, f f(x).
Proof. Calculate according to 2.1.1 (f + g) (f + g)(x) = f(x) + g(x) and
(af) (af)(x) = af(x).
2.5. HOMOMORPHISM MODULES 33
2.5.10. Denition. The natural homomorphism 2.5.9
ev
x
: Hom
R
(M, N) N, f f(x)
is the evaluation at x.
2.5.11. Lemma. There is a natural homomorphism
M Hom
R
(Hom
R
(M, N), N), x ev
x
Proof. Calculate according to 2.1.1 ev
x+y
(f) = f(x + y) = f(x) + f(y) =
ev
x
(f) + ev
y
(f). and ev
ax
(f) = f(ax) = af(x) = aev
x
(f) to see that the map
is a homomorphism.
2.5.12. Proposition. Let R be a ring and M a module. The evaluation
ev
1
: Hom
R
(R, M) M, f f(1)
is a natural isomorphism. x 1
x
2.3.11 is the inverse.
Proof. Calculate the composite ev
1
(x 1
x
) = 1
x
(1) = x and 1
f(1)
(a) =
af(1) = f(a) proving the claims.
2.5.13. Denition. Let R be a ring and M a module. The dual module is
M

= Hom
R
(M, R)
If f : M N is a homomorphism, then the dual homomorphism is
f

= Hom(f, 1
R
) : N

This construction is a contravariant functor.


2.5.14. Lemma. There is a natural homomorphism
M M

= Hom
R
(Hom
R
(M, R), R), x ev
x
where ev
x
(f) = f(x) 2.5.10.
Proof. This is a special case of 2.5.11
2.5.15. Denition. Amodule M is a reexive module if the homomorphism2.1.14,
M M

is an isomorphism.
2.5.16. Example. Let R be a ring.
(1) The module R is reexive.
(2) If (a) ,= R, (0) in a domain, then R/(a) is not reexive.
2.5.17. Exercise. (1) Show that Hom
Z
(Q, Z) = 0.
(2) Calculate Hom
Z
(Z/(m), Z/(n)) = 0 for integers m, n.
(3) Let I R be an ideal and M a module. Show that Hom
R
(R/I, M) = x M[I
Ann(x).
(4) Let R be a ring. Show that a free module with a nite basis is a reexive module.
(5) If (n) (m) Z, then show that (m)/(n) is a reexive Z/(n)-module.
34 2. MODULES
2.6. Tensor product modules
2.6.1. Denition. Let R be a ring and M, N modules. The tensor product module
M
R
N = F/F

is the factor module F/F

, where F =
MN
R is the free module with basis
(x, y) = e
(x,y)
2.4.6 and F

is the submodule generated by all elements of form


(x
1
+x
2
, y) (x
1
, y) (x
2
, y), (x, y
1
+y
2
) (x, y
1
) (x, y
2
)
(ax, y) a(x, y), (x, ay) a(x, y)
The projection of the basis element (x, y) onto M
R
N is x y = (x, y) +F

.
2.6.2. Remark. The relations are interpreted.
(1) There are identities in M
R
N
(x
1
+x
2
) y = x
1
y +x
2
y, x (y
1
+y
2
) = x y
1
+x y
2
ax y = a(x y) = x ay
(2) The map
: M N M
R
N, (x, y) x y
has partial maps x xy : M M
R
N and y xy : N M
R
N
that are all homomorphisms.
(3) The formation of partial homomorphism are again homomorphisms.
N Hom
R
(M, M
R
N)), y (x x y)
and
M Hom
R
(N, M
R
N)), x (y x y)
2.6.3. Theorem. Given a map : M N L such that the partial maps x
(x, y) : M L and y (x, y) : N L are homomorphisms. Then there
exists a unique homomorphism u : M
R
N L such that u(x y) = (x, y).
M N


M
R
N
u

L
On a general element u satises
u(

) =

u(x

) =

(x

, y

)
Proof. By 2.6.1 M
R
N = F/F

. The homomorphism 2.4.11 F L, (x, y)


(x, y) has F

in the kernel. 2.3.5 gives the homomorphism u.


2.6.4. Remark. Assume x

generate M and y

generate N.
(1) x

generate M
R
N.
(2) Two homomorphisms u, v : M
R
N L are equal if u(x

) =
v(x

).
2.6.5. Proposition. Let R be a ring and f : M M

, g : N N

homomor-
phisms of modules. Then there is induced a homomorphism
M
R
N M

R
N

, x y f(x) g(y)
2.6. TENSOR PRODUCT MODULES 35
Proof. The map south-east
M N
fg


M
R
N

R
N

satises the assumptions in 2.6.3 to induce the right vertical map satisfying xy
f(x) g(y).
2.6.6. Denition. f g : M
R
N M

R
N

is the induced homomorphism


2.6.5. On a general element
f g(

) =

f(x

) g(y

)
2.6.7. Proposition. Let R be a ring. The constructions
(1)
M M
R
N, f f 1
N
(2)
N M
R
N, g 1
M
g
are functors and
Hom
R
(M, M

) Hom
R
(M
R
N, M

R
N)
Hom
R
(N, N

) Hom
R
(M
R
N, M
R
N

)
are R-homomorphism.
Proof. Given also homomorphisms f

: M

and g

: N

. Then by
2.6.4
f

f g

g = f

f g
The rest follows directly from 2.6.5. E.g.
(f +g) 1 = f 1 +g 1, (af) 1 = a(f 1)
and similarly.
2.6.8. Corollary. Let a R give scalar multiplications a
M
, a
N
. The homomor-
phisms
a
M
1
N
= 1
M
a
N
= a
M
R
N
are the same homomorphism
M
R
N M
R
N, x y a(x y)
2.6.9. Example. Let R be a ring.
(1) Then there is an isomorphism
R
R
R R, a b ab
(2) Let M, N, L be modules. Composition of maps gives a homomorphism
Hom
R
(N, L)
R
Hom
R
(M, N) Hom
R
(M, L), g f g f
by 2.5.3. This is a natural homomorphism in each variable.
36 2. MODULES
(3) For a module M composition
Hom
R
(M, M)
R
Hom
R
(M, M) Hom
R
(M, M), g f g f
gives Hom
R
(M, M) a structure of a normally noncommutative ring. The
map R Hom
R
(M, M), a a
M
is a ring homomorphism.
2.6.10. Proposition. Let Rbe a ring and M, N, Lmodules. Then there are natural
isomorphisms
(1) (unit:) M
R
R M, x a ax
(2) (commutative:) (M
R
N) N
R
M, x y y x
(3) (associative:) (M
R
N)
R
L M
R
(N
R
L), (xy)z x(yz)
Proof. (1) M R M, (x, a) ax induces the homomorphism M
R
R
M, x a ax by 2.6.3. The map M R
R
M, x 1 x is the inverse. (2)
M N N
R
M, (x, y) y x induces the homomorphism (M
R
N)
N
R
M, x y y x by 2.6.3. The inverse is constructed similarly and the
composites are the identities by the uniqueness statement in 2.6.3. (3) For a xed
z L the map M N M
R
(N
R
L), (x, y) x (y z) induces the
homomorphism
z
: M
R
N M
R
(N
R
L), xy x(y z) by 2.6.3.
Finally the map (M
R
N)L M
R
(N
R
L), (xy, z)
z
(xy). induces
the homomorphism (M
R
N)
R
L M
R
(N
R
L), (xy)z x(yz)
by 2.6.3. The inverse is constructed similarly and the composites are the identities
by the uniqueness statement in 2.6.3.
2.6.11. Proposition. Let Rbe a ring and M

a family of modules. For any module


N there is a natural isomorphism
(

)
R
N

(M

R
N)
giving the identication (

) y =

(x

y).
Proof. 2.4.3 and 2.6.3 give a pair of inverse homomorphisms. Fix y N. The
family g

: M



(M

R
N), x

y induces by 2.4.3 a homomor-


phism g
y
:

M



(M


R
N). The map (

) N

(M


R
N), (

, y) g
y
(

) induces by 2.6.3 the homomorphism (

)
R
N

(M

R
N), (

)y

y. The family i

1
N
: M

R
N
(

)
R
N induces by 2.4.3 the inverse.
2.6.12. Example. Let R be a ring, F a free module with basis y

and G a free
module with basis z

. Then F
R
G is a free module with basis y

.
2.6.13. Theorem. Let R be a ring and M, N, L modules. Then there is a natural
isomorphism
Hom
R
(M
R
N, L) Hom
R
(M, Hom
R
(N, L))
f (x [y f(x y)])
(x y g(x)(y)) g
Proof. A given f : M
R
N L is mapped to the composite homomorphism
M Hom
R
(N, M
R
N) Hom
R
(N, L), 2.5.4 and 2.6.2. This is a homomor-
phism as map of f by 2.5.4. Given g : M Hom
R
(N, L) the map M N
L, (x, y) g(x)(y) induces a homomorphism M
R
N L, x y g(x)(y)
2.7. CHANGE OF RINGS 37
by 2.6.3. Clearly the maps are inverse to each other and therefore giving an iso-
morphism by 2.1.5.
2.6.14. Exercise. (1) Show that Q
Z
Q/Z 0.
(2) Show that Z/(m)
Z
Z/(n) = 0 if (m, n) = Z.
(3) Let P, Q R be different maximal ideals and M a module. Show that M/PM
R
M/QM = 0.
2.7. Change of rings
2.7.1. Proposition. (1) Let : R S be a ring homomorphism and N an S
module. The restriction scalars 2.1.11 viewing N as an R-module through
, R N N, (a, x) ax = (a)x, is a functor from S-modules to
R-modules.
(2) Let I R be an ideal. Restriction of scalars along R R/I identies
R/I-modules M with R-modules such that I Ann(M). For R/I-modules
M, N there is a natural isomorphism
Hom
R
(M, N) Hom
R/I
(M, N)
Proof. This is a restatement of 2.1.11 using 2.5.4.
2.7.2. Lemma. Let R S be a ring homomorphism, M an R-module and N an
S-module. Then
S M
R
N M
R
N, (b, x y) x by
is an S-scalar multiplication.
Proof. For xed b S the map M N M
R
N, (x, y) x by induces
the homomorphism
b
: M
R
N M
R
N, x y x by by 2.6.3. This
gives a well dened scalar multiplication S M
R
N M
R
N, (b, xy)

b
(x y).
2.7.3. Denition. Let R S be a ring homomorphism and M an R-module. The
change of ring S-module is M
R
S with S-scalar multiplication 2.7.2
S M
R
S M
R
S, (b, x c) x bc
2.7.4. Proposition. The construction
M M
R
S
and
f : M M

f 1
S
: M
R
S M

R
S
is a functor from R-modules to S-modules.
Hom
R
(M, M

) Hom
S
(M
R
S, M

R
S)
is an R-homomorphism.
Proof. By 2.6.7 this is a functor to R-modules and by 2.7.2 f 1
S
is an S-
homomorphism.
2.7.5. Proposition. Let R S be a ring homomorphism, M, M

two R-modules
and N an S-modules.
38 2. MODULES
(1) There is a natural isomorphism of S-modules.
M
R
S
S
N M
R
N
x b y x by
(2) Then there is a natural isomorphism of S-modules.
(M
R
S)
S
(M

R
S) (M
R
M

)
R
S
x b y c x y bc
Proof. (1) The homomorphism v : S
S
N N, b y by is an isomorphism,
2.6.10. The homomorphism 1
M
v : M
R
S
S
N M
R
N is an R-module
isomorphism, 2.6.7. The identity x bc y = x b cy proves this to be an
S-module homomorphism. (2) Use (1) on N = M


R
S and conclude by the
canonical isomorphisms 2.6.10.
2.7.6. Theorem. Let R S be a ring homomorphism, M, M

two R-modules
and N an S-modules.
(1) For any R-homomorphism f : M N, then there exists a unique S-
homomorphism f

: M
R
S N, x b bf(x) such that f(x) =
f

(x 1).
M

N
M
R
S
f

(2) There is a natural isomorphism


Hom
R
(M, N) Hom
S
(M
R
S, N)
f (x b bf(x))
(3) There is a natural homomorphism
Hom
R
(M, N)
R
S Hom
S
(M
R
S, M

R
S)
f c (x b f(x) bc)
Proof. (1) From 2.6.3 f

is the unique R-homomorphism and by 2.7.3 this is an


S-homomorphism. (2) This is a restatement of (1). A given f is mapped to the
composite M
R
S N
R
S N which is an S-homomorphism. Given a
homomorphism g : M
R
S N then the composite M M
R
S N is an
R-homomorphism and an inverse to the rst given map. (3) This follows from (1)
applied to Hom
R
(M, N) Hom
S
(M
R
S, M

R
S).
2.7.7. Lemma. Let R S be a ring homomorphism, M an R-module and N an
S-module. Then
S Hom
R
(N, M) Hom
R
(N, M)
(b, f : N M) (y f(by))
is an S-scalar multiplication.
Proof. The map (b, f) f b
N
satises the laws 2.1.1.
2.7. CHANGE OF RINGS 39
2.7.8. Denition. Let R S be a ring homomorphism and M an R-module. The
induced module is the S-module Hom
R
(S, M) with S-scalar multiplication 2.7.7
S Hom
R
(S, M) Hom
R
(S, M)
(b, f : S M) (c f(bc))
2.7.9. Proposition. The induced module
M Hom
R
(S, M)
and
f : M M

Hom(1
S
, f) : Hom
R
(S, M) Hom
R
(S, M

)
is a functor from R-modules to S-modules.
Hom
R
(M, M

) Hom
S
(Hom
R
(S, M), Hom
R
(S, M

))
is an R-homomorphism.
Proof. By 2.5.5 this is a functor to R-modules and by 2.7.7 Hom(1, f) is an S-
homomorphism.
2.7.10. Theorem. Let R S be a ring homomorphism and M an R-module and
N an S-modules. Then there is a natural isomorphism
Hom
R
(N, M) Hom
S
(N, Hom
R
(S, M))
f (y [b f(by)])
Proof. g (y g(y)(1)) is an inverse.
2.7.11. Example. Let I R be an ideal and R R/I the projection.
(1) The change of ring functor maps an R-module M to the R/I-module M/IM.
The natural isomorphism 2.7.6 is
Hom
R
(M, N) Hom
R/I
(M/IM, N)
for any R/I-module N.
(2) The induced module functor maps an R-module M to the R/I-module x
M[I Ann(x). The natural isomorphism 2.7.10 is
Hom
R
(N, M) Hom
R/I
(N, Hom
R
(R/I, M))
for any R/I-module N.
2.7.12. Denition. Let R S, S

be ring homomorphisms. The tensor product


ring over R is S
R
S

with multiplication given by (b b

)(c c

) = bc b

extended by linearity. R S
R
S

, r r 1 = 1 r is the natural ring


homomorphism.
2.7.13. Proposition. Let ,

: R S, S

and ,

: S, S

T give a commu-
tative diagram of ring homomorphisms, =

. Then b b

(b)

(b

)
40 2. MODULES
is the unique homomorphism making the following diagram commutative.
R

S
R
S

T
Proof. This is clear by 2.6.3.
2.7.14. Example. Let R S be a ring homomorphism. Then
R[X]
R
S S[X]
is an isomorphism.
2.7.15. Exercise. (1) Show that the change of rings of a free R-module is a free S-
module.
(2) Let : R S be a ring homomorphism. Show that the change of rings of a
scalar multiplication a : M M on an R-module is a scalar multiplication (a) :
M
R
S M
R
S.
(3) Show that the change of rings of the composition of two homomorphisms is the
composition of the change of rings of each homomorphism.
(4) Show the isomorphism
R[X]
R
R[Y ] R[X, Y ]
3
Exact sequences of modules
3.1. Exact sequences
3.1.1. Denition. Let f : M N and g : N L be homomorphisms of
modules. The sequence
M
f

N
g

L
of homomorphisms is a
(1) 0-sequence: g f = 0 or equivalently Imf Ker g
(2) exact sequence: Imf = Ker g
For a sequence of more homomorphisms the conditions should be satised for
every consecutive composition. E.g. The sequence
M
f

N
g

L
h

K
is a 0-sequence if g f = 0 and h g = 0. The sequence is exact if Imf = Ker g
and Img = Ker h.
3.1.2. Remark. An interpretation of 2.3.3 gives:
(1) The sequence
0

M
f

N
is exact if and only if f is injective.
(2) The sequence
M
f

N

0
is exact if and only if f is surjective.
(3) The sequence
0

M
f

N

0
is exact if and only if f is an isomorphism.
3.1.3. Proposition. (1) For a homomorphism f : M N the sequence
0

Ker f

M
f

N

Cok f

0
is exact.
(2) For scalar multiplication with a R on M the sequence
0

Ker a
M

M
a
M

M

M/aM

0
is exact.
41
42 3. EXACT SEQUENCES OF MODULES
3.1.4. Proposition. Given a 0-sequence
0

M
f

N
g

L
then the following statements are equivalent:
(1) The sequence is exact.
(2) f is an isomorphism onto Ker g.
(3) Given a homomorphism h : K N such that g h = 0 then there is a
unique h

: K M such that h = f h

.
0

M
f

N
g

L
K
h

Proof. (1) (2): This is clear. (1) (3): Imh Ker g = Imf, so by (2) put
h

= f
1
h. (3) (2): Apply it to Ker g M to see that (2) is satised.
3.1.5. Proposition. Given a 0-sequence
M
f

N
g

L

0
then the following statements are equivalent:
(1) The sequence is exact.
(2) The factor homomorphism 2.3.9 g

: Cok f L induced by g is an isomor-


phism.
(3) Given a homomorphism k : N K such that k f = 0 then there is a
unique k

: L K such that k = k

g.
M
f

N
g

L
k

0
K
Proof. (1) (2): This follows from 2.3.9. (1) (3): By 2.3.5 there is k

:
Cok f K such that k

p = k. By (1) put k

= k

g
1
. (3) (2) satised
and apply it to N Cok f to see that (2) is satised.
3.1.6. Proposition. Let
M

be a family of exact sequences. Then there are exact sequences:


(1) The sum

(2) The product

Proof. Clear since kernel and image are calculated componentwise.


3.1. EXACT SEQUENCES 43
3.1.7. Denition. An exact sequence
0

M
f

N
g

L

0
is a short exact sequence. That is f is injective, Imf = Ker g and g is surjective.
3.1.8. Proposition. (1) Let I R be an ideal, then there is a short exact se-
quence
0

I

R

R/I

0
(2) Let M N be a submodule, then there is a short exact sequence
0

M

N

N/M

0
(3) For scalar multiplication with nonzero divisor a R on M the sequence
0

M
a
M

M

M/aM

0
is a short exact sequence.
(4) Given a homomorphism f : M N there are associated two short exact
sequences.
0

Ker f

M
f

Imf

0
and
0

Imf

N

Cok f

0
(5) For scalar multiplication with any a Ron M there are associated two short
exact sequences.
0

Ker a
M

M
a
M

aM

0
and
0

aM

M

M/aM

0
3.1.9. Denition. Let f : M N be a homomorphism.
(1) f has a retraction if there is a homomorphism u : N M such that u f =
1
M
.
(2) f has a section if there is a homomorphismv : N M such that f v = 1
N
.
3.1.10. Proposition. Let f : M N be a homomorphism.
(1) If f has a retraction u : N M then f is injective, u is surjective and
N = Imf Ker u
(2) If f has a section v : N M then f is surjective, v is injective and
M = Ker f Imv
Proof. (1) u(f(x)) = x so f is injective and u is surjective. If y N then
y = f(u(y)) +(y f(u(y)) and u(y f(u(y))) = 0, so N = Imf +Ker u. Let
y Imf Ker u. Then y = f(x) gives x = u(f(x)) = u(y) = 0, so y = 0.
Conclude by 2.4.5 that the sum is direct. (2) y = f(v(y)) so f is a retraction of v.
Finish by (1).
44 3. EXACT SEQUENCES OF MODULES
3.1.11. Lemma. For a short exact sequence
0

M
f

N
g

L

0
the following statements are equivalent:
(1) f has a retraction.
(2) g has a section.
For any retraction u there is a unique section v and wise-verse such that
1
N
= f u +v g
Proof. If u is a retraction of f, then Ker g = Imf Ker(1
N
f u). By 3.1.5
there is a homomorphism v : L N such that v g = 1
N
f u. This is a
section of g. Conversely if v is a section of g then Im(1
N
v g) Ker g, so
there is a homomorphism u : N M such that f u = 1
N
v g, 3.1.4. u is a
retraction of f. The equation is clearly satised.
3.1.12. Denition. Let R be a ring and f : M N, g : N L homomorphisms.
A short exact sequence
0

M
f

N
g

L

0
is a split exact sequence if equivalently 3.1.11 f has a retraction or g has a section.
3.1.13. Theorem. A sequence
0

M
f

N
g

L

0
is a split exact sequence if and only if there are homomorphism u : N M, v :
L N satisfying
u f = 1
M
, g v = 1
L
, f u +v g = 1
N
If the sequence is split exact then
0

L
v

N
u

M

0
is split exact and (x, y) f(x)+v(y) and z u(z)+g(z) gives the isomorphism
M L N
Proof. The sequence is a 0-sequence f is injective and g is surjective. From f
u+v g = 1
N
follows that z Ker g Imf, so the sequence is short exact. The
rest is contained in 3.1.10.
3.1.14. Corollary. A (contravariant) functor preserves split exact sequences. If
0

M
f

N
g

L

0
is split exact and T a functor, then
0

T(M)
T(f)

T(N)
T(g)

T(L)

0
is split exact. If T is contravariant, then
0

T(L)
T(g)

T(N)
T(f)

T(M)

0
is split exact.
3.2. THE SNAKE LEMMA 45
Proof. By 3.1.13 a split exact sequence is characterized by a set of equations.
These are preserved by the functor, 2.5.4.
3.1.15. Example. A short exact sequence
0

M
f

N
g

F

0
where F is a free module is a split exact sequence. Namely let x

F be a
basis and choose y

N with g(y

) = x

. The dene v : L N by setting


v(x

) = y

, 2.4.11.
3.1.16. Example. Let Z
i
be the family of modules each a copy of Z indexed by
the natural numbers. Then the short exact sequence
0

i
Z
i

i
Z
i

i
Z
i
/

i
Z
i

0
is not split exact.
The element f = (1, 2, 2
2
, . . . , 2
n
, . . . ) +

i
Z
i
is divisible by 2
k
for all k. If
f
k
= (0, . . . , 0, 2
nk
, . . . ) +

i
Z
i
for n k, then 2
k
f
k
= f in

i
Z
i
/

i
Z
i
.
But in

i
Z
i
the only element divisible with all 2
k
is 0, so no section exists.
3.1.17. Exercise. (1) Show that the sequence
0

Z

Q

Q/Z

0
is short exact, but not split exact.
(2) Show that the sequence
0

Z
n

Z

Z/(n)

0
is exact, but not split exact for n ,= 0, 1.
(3) Show that the sequence
0

Z/(2)
12

Z/(4)

Z/(2)

0
is exact, but not split exact.
(4) Show that the sequence
0

Z/(2)
13

Z/(6)

Z/(3)

0
is split exact.
3.2. The snake lemma
3.2.1. Example. Given a commutative diagram of homomorphisms
M
u

N
v

there is induced a commutative diagram


0

Ker f
u

M
u

N
v

Cok f
v

0
0

Ker f

M

N

Cok f

0
where the rows are exact sequences.
46 3. EXACT SEQUENCES OF MODULES
The diagram splits into two diagrams
0

Ker f
u

M
u

Imf
v

0
0

Ker f

M

Imf

0
and
0

Imf
v

N
v

Cok f
v

0
0

Imf

N

Cok f

0
where the rows are short exact sequences.
3.2.2. Lemma. Given a commutative diagram of homomorphisms
M
u

N
v

L
w

0
0

M

where the rows exact sequences. The snake homomorphism : Ker w Cok u
is well dened by: For z Ker w choose y N such that g(y) = z. The
element v(y) Ker g

so there is x

such that f

(x

) = v(y). Then (z) =


x

+ Imu Cok u.
Proof. Assume g(y

) = z and f

(x

) = v(y

). There is x M with f(x) = yy

.
Now f

(u(x)) = v(f(x)) = v(y y

) = f

(x

) so u(x) = x

since f

is injective. Then x

+ Imu = x

+ Imu as wanted. The choices made respect


addition and scalar multiplication showing that is a homomorphism.
3.2.3. Remark. The snake is
Ker u

Ker v

Ker w


M
u

N
v

L
w

0
0

M

Cok u
f

Cok v
g

Cok w
3.2. THE SNAKE LEMMA 47
The construction of is schematically
Ker w

N
v

L
M

Cok u
z


z
x


v(y)
(z)
3.2.4. Theorem (snake lemma). Given a commutative diagramof homomorphisms
M
u

N
v

L
w

0
0

M

where the rows exact sequences. There is induced a six term long exact sequence
Ker u
f

Ker v
g

Ker w

Cok u
f

Cok v
g

Cok w
Proof. By construction of it is clear that the sequence is a 0-sequence: If y
Ker v then to calculate (g(y)) the choice v(y) = 0 gives g = 0. Also
f

((z)) = v(y) + Imv shows that f

= 0. Exactness at Ker v and Cok v


are clear. Given z Ker w such that (z) = 0. By 3.2.2 choose y, g(y) = z
and x

, f

(x

) = v(y). Then (z) = x

+ Imu = 0, so choose x, u(x) = x

.
Now v(f(x)) = f

(u(x)) = v(y) so y f(x) Ker v and g(y f(x)) =


g(y) = z. Therefore exactness at Ker w. Given x

+ Imu Cok u such that


f

(x

) + Imv = 0 Cok v. Choose y, v(y) = f

(x

) and put z = g(y). Then


w(g(y)) = g

(v(y)) = g

(f

(x

)) = 0. Now z Ker w and (z) = x

+ Imu.
Therefore exactness at Cok u.
3.2.5. Corollary. If f is injective then the f : Ker u Ker v is injective and the
long exact sequence is
0

Ker u
f

Ker v
g

Ker w

Cok u
f

Cok v
g

Cok w
If g

is surjective then g

: Cok v Cok w is surjective and the long exact se-


quence is
Ker u
f

Ker v
g

Ker w

Cok u
f

Cok v
g

Cok w

0
48 3. EXACT SEQUENCES OF MODULES
3.2.6. Corollary. (1) If v is injective and u is surjective, then w is injective.
(2) If v is surjective and w is injective, than u is surjective.
(3) If v is an isomorphism, then w is injective if and only if u is surjective.
3.2.7. Proposition. Given submodules N, L M, then there is a short exact
sequence
0

M/N L
x(x,x)

M/N M/L
(x,y)xy

M/N +L

0
Proof. The commutative diagram
0

N L

N L

N +L

0
0

M
x(x,x)

M M
(x,y)xy

M

0
where the rows are short exact sequences, gives by 3.2.4 a ve term long exact
sequence
0

M/N L

M/N M/L

M/N +L

0
3.2.8. Proposition (ve lemma). Given a commutative diagramof homomorphisms
M
1
u
1

M
2
u
2

M
3
u
3

M
4
u
4

M
5
u
5

1

M

2

M

3

M

4

M

5
where the rows are exact sequences. If u
1
is surjective, u
2
, u
4
are isomorphism
and u
5
is injective, then u
3
is an isomorphism.
Proof. Let f
i
: M
i
M
i+1
, f

i
: M

i
M

i+1
. Split the given diagram in three as
follows
M
1
u
1

M
2
u
2

Cok f
1
u

0
0

Ker f

2

M

2

Imf

2

0
0

Cok f
2
u

M
4
u
4

Imf
4
u
5

0
0

Ker f

4

M

4

M

5
0

Imf
2
u

M
3
u
3

Cok f
2
u

0
0

Imf

2

M

3

Cok f

2

0
Note that Cok f
1
Imf
2
and Cok f

2
Ker f

4
. Now use 2.3.3 and the snake
lemma to conclude that Ker u
3
= 0 and Cok u
3
= 0 and u
3
is therefore an isomor-
phism.
3.2. THE SNAKE LEMMA 49
3.2.9. Corollary. (1) Given a commutative diagram of homomorphisms
M
u

N
v

If u, v are isomorphism, then the induced homomorphisms u : Ker f


Ker f

and v : Cok f Cok f

are isomorphisms.
(2) Given a commutative diagram of homomorphisms
M
u

N
v

L
w

M

N

L

If u, v, w are isomorphism, then the upper row is exact if and only if the lower
row is exact.
3.2.10. Proposition (windmill lemma). Given homomorphism
M
f

N
g

L
.
There is induced an eight term long exact sequence
0

Ker f

Ker g f
f

Ker g

Cok f
g

Cok g f

Cok g

0
Proof. Look at the two diagrams
0

M
f

M
gf

0
0

Ker g

N
g

L
M
gf

N
g

Cok f

0
0

L
1

L

0
By the snake lemma the sequences
0

Ker f

Ker g f

=f

Ker g

Cok f

Cok g f
Ker g f

Ker g

Cok f

=g

Cok g f

Cok g

0
are exact and overlap to give the windmill sequence.
50 3. EXACT SEQUENCES OF MODULES
3.2.11. Remark. The windmill is
Ker g f

Ker g

Ker f

M
f

gf

N

g

Cok f
/

..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Cok g
_

Cok g f

3.2.12. Exercise. (1) Given a short exact sequence


0

M

N

L

0
Show that Ann(N) Ann(M) Ann(L).
(2) Give a short exact sequence
0

M

N

L

0
such that Ann(N) ,= Ann(M) Ann(L).
(3) Given ideals I, J R. Show that there is a short exact sequence.
0

R/I J

R/I R/J

R/I +J

0
3.3. Exactness of Hom
3.3.1. Denition. A functor T is a left exact functor if given an exact sequence
0

M
f

N
g

L
the following sequence is exact
0

T(M)

T(N)

T(L)
T is a right exact functor if given an exact sequence
M
f

N
g

L

0
the following sequence is exact
T(M)

T(N)

T(L)

0
A functor is an exact functor if it is both left and right exact.
A contravariant functor T is a left exact contravariant functor if given an exact
sequence
0

M
f

N
g

L
the following sequence is exact
T(L)

T(N)

T(M)

0
T is a right exact contravariant functor if given an exact sequence
M
f

N
g

L

0
3.3. EXACTNESS OF HOM 51
the following sequence is exact
0

T(L)

T(N)

T(M)
A contravariant functor is an exact functor if it is both left and right exact.
3.3.2. Proposition. Let T be a functor. The following conditions are equivalent:
(1) T is a exact.
(2) Given an exact sequence
M
f

N
g

L
then the following sequence is exact
T(M)

T(N)

T(L)
(3) Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is short exact
0

T(M)

T(N)

T(L)

0
Let T be a contravariant functor. The following conditions are equivalent:
(1) T is a exact.
(2) Given an exact sequence
M
f

N
g

L
then the following sequence is exact
T(L)

T(N)

T(M)
(3) Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is short exact
0

T(L)

T(N)

T(M)

0
Proof. Let T be a functor. (1) (2): The exact sequence M Imf 0 gives
T(M) T(Imf) 0 exact and the exact sequence 0 Imf N gives
0 T(Imf) T(N) exact. It follows that ImT(f) T(Imf). The sequence
0 Imf N L is exact so 0 T(Imf) T(N) T(L) is exact. Now
conclude that ImT(f) = Ker T(g). (2) (3): This is clear. (3) (1): Given
0
M
f

N
g

L
exact, then the sequences 0 T(M) T(N)
T(Img) 0 and 0 T(Img) T(L) are exact, so ImT(f) = ker T(g) and T
is left exact. Similarly T is rigth exact.
3.3.3. Corollary. Let T be an exact functor. For a homomorphism f : M N
there are natural isomorphisms.
(1) Ker T(f) T(Ker f).
(2) ImT(f) T(Imf).
(3) Cok T(f) T(Cok f).
52 3. EXACT SEQUENCES OF MODULES
Let T be a contravariant functor. For a homomorphism f : M N there are
natural isomorphisms.
(1) Ker T(f) T(Cok f).
(2) ImT(f) T(Imf).
(3) Cok T(f) T(Ker f).
Proof. Represent the statements using short exact sequences. (1) The kernel is
determined by the exact sequence 0 Ker f M N, 3.1.4. (3) The cokernel
is determined by the exact sequence M N Cok f 0, 3.1.5. (2) The
image is determined by the exact sequence 0 Ker f M Imf 0, 2.3.5,
3.1.4.
3.3.4. Corollary. Let T be an exact functor. For submodules N, L M there are
natural identications
(1) T(M/N) = T(M)/T(N).
(2) T(N +L) = T(N) +T(L).
(3) T(N L) = T(N) T(L).
Proof. Represent the statements using short exact sequences. (1) 0 N M
M/N 0 is short exact giving 0 T(N) T(M) T(M/N) 0. The
wanted isomorphism follows form 3.1.5. (2) N +L is the image of N L M.
(3) N L is the kernel of N L M.
3.3.5. Proposition. The functor Hom
R
(K, ) is left exact. Given an exact se-
quence
0

M
f

N
g

L
and an R-module K. Then the following sequence is exact
0

Hom
R
(K, M)

Hom
R
(K, N)

Hom
R
(K, L)
Proof. Given h : K M such that f h = 0 then h = 0 since f is injective.
Given k : K N such that g k = 0 then by 3.1.4 there is h : K M such that
f h = k. So the sequence is exact.
3.3.6. Proposition. The contravariant functor Hom
R
(, K) is right exact. Given
an exact sequence
M
f

N
g

L

0
and a module K. Then the following sequence is exact
0

Hom
R
(L, K)

Hom
R
(N, K)

Hom
R
(M, K)
Proof. Given h : L K such that h g = 0 then h = 0 since g is surjective.
Given k : N K such that k f = 0 then by 3.1.5 there is h : L K such that
h g = k. So the sequence is exact.
3.3.7. Proposition. Given a sequence
M
f

N
g

L

0
such that for any module K, the following sequence is exact
0

Hom
R
(L, K)

Hom
R
(N, K)

Hom
R
(M, K)
Then the original sequence is exact.
3.4. EXACTNESS OF TENSOR 53
Proof. For K = L the identity 1
L
is mapped to 1
L
g f = 0 so it is a 0-sequence.
Take K = Cok g then p
g
: L Cok g has p
g
g = 0, but by exactness 0 is the
unique homomorphism satisfying this, so p
g
= 0. Therefore Cok g = 0 and g is
surjective. Take K = Cok f, p : N Cok f the projection. p f = 0 so by
exactness there exists a unique q : L Cok f such that q g = p. It follows that
Ker g Ker p = Imf. All together the original sequence is exact.
3.3.8. Proposition. Given a sequence
0

M
f

N
g

L

0
The following conditions are equivalent:
(1) The sequence is split exact.
(2) For any K the following sequence is exact
0

Hom
R
(K, M)

Hom
R
(K, N)

Hom
R
(K, L)

0
(3) For any K the following sequence is exact
0

Hom
R
(L, K)

Hom
R
(N, K)

Hom
R
(M, K)

0
If the conditions are true, then the sequences (2) and (3) are split exact.
Proof. (1) (2), (1) (3) are clear by 3.1.14 giving that the sequences (2), (3)
are split exact. (2) (1): Let K = L, then there is a section to g. By 3.1.11 the
original sequence is split exact. (3) (1): Let K = M, then there is a retraction
to f. By 3.3.7and 3.1.11 the original sequence is split exact.
3.3.9. Exercise. (1) Show that the sequence
0

Hom
Z
(Q/Z, Z)

Hom
Z
(Q/Z, Q)

Hom
Z
(Q/Z, Q/Z)
is exact, but the rightmost map is not surjective.
(2) Show that the sequence
0

Hom
Z
(Z/(n), Z)
n

Hom
Z
(Z/(n), Z)

Hom
Z
(Z/(n), Z/(n))
is exact, but the rightmost map is not surjective.
3.4. Exactness of tensor
3.4.1. Theorem. The functor
R
K is right exact. Given an exact sequence
M
f

N
g

L

0
and an R-module K. Then the following sequence is exact
M
R
K

N
R
K

L
R
K

0
Proof. Let K

be any module. By 3.3.7 it is enough to see that the sequence


0

Hom
R
(L
R
K, K

)

Hom
R
(N
R
K, K

Hom
R
(M
R
K, K

)
54 3. EXACT SEQUENCES OF MODULES
is exact. By 2.6.13 it amounts to see that the sequence
0

Hom
R
(L, Hom
R
(K, K

))

Hom
R
(M, Hom
R
(K, K

))

Hom
R
(N, Hom
R
(K, K

))
is exact. This follows from 3.3.6
3.4.2. Proposition. Given a split exact sequence
0

M
f

N
g

L

0
and a module K. Then the following sequence is split exact
0

K
R
M

K
R
N

K
R
L

0
Proof. This follows from the functor properties 3.1.14.
3.4.3. Proposition. Let I Rbe an ideal. For any module M, the homomorphism
M
R
R/I M/IM, x a +I ax +IM
is an isomorphism.
Proof. There is a commutative diagrm
M
R
I

M
R
R

M
R
(R/I)

0
0

IM

M

M/IM

0
The conclusion follows from the snake lemma, 3.2.4.
3.4.4. Corollary. Let I, J R be ideals. Then
R/I
R
R/J R/(I +J), a +I b +J ab +I +J
is an isomorphism.
3.4.5. Proposition. Let I
1
, . . . , I
k
R be pairwise comaximal ideals and M a
module. Then the product of projections
M/I
1
I
k
M M/I
1
M M/I
k
M
is an isomorphism.
Proof. By Chinese remainders 1.4.2
R/I
1
I
k
R/I
1
R/I
k
is an isomorphism. Tensor with M and use 3.4.3 to get the isomorphism.
3.4.6. Exercise. (1) Calculate Z/(m)
Z
Z/(n) for all integers m, n.
(2) Let I R be an ideal. Show that R/I
R
R/I R/I.
(3) Let I R be an ideal. Show that I
R
R/I I/I
2
.
(4) Let 2
Z
be scalar multiplication. Show that 2
Z
1
Z/(2)
: Z
Z
Z/(2) Z
Z
Z/(2)
is not injective.
3.5. PROJECTIVE MODULES 55
3.5. Projective modules
3.5.1. Denition. An R-module F is a projective module if for any surjectice
homomorphism N L the homomorphism
Hom
R
(F, N) Hom
R
(F, L)
is surjective.
3.5.2. Proposition. Let F be an R-module. The following conditions are equiva-
lent:
(1) F is projective.
(2) The functor Hom
R
(F, ) is exact.
(3) Given an exact sequence
M

N

L
then the following sequence is exact
Hom
R
(F, M)

Hom
R
(F, N)

Hom
R
(F, L)
(4) Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is short exact
0

Hom
R
(F, M)

Hom
R
(F, N)

Hom
R
(F, L)

0
Proof. This is clear from 3.3.2. (1) (2): This follows since Hom
R
(F, ) is
allways left exact. (1) (3) (4): This is true for any exact functor.
3.5.3. Proposition. A module F is projective if and only if any surjective homo-
morphism M F 0 has a section.
Proof. Assume F projective and g : M F surjective. Then Hom
R
(F, M)
Hom
R
(F, F) 0 is exact. So there exists a v : F M such that g v = 1
F
.
v is then a section. Conversely given g : N L surjective and h : F L. Let
M = Ker NF L, (y, z) g(y)h(z) and p
N
: M N, p
F
: M F the
projections, then g p
N
= hp
F
. Now p
F
is surjective since g is. Let v : F M
be a section of p
F
, then h

= p
N
v satises h = g h

.
N
g

L

0
M
p
N

p
F

F
h

v
.
h

3.5.4. Corollary. A short exact sequence


0

M
f

N
g

F

0
where F is a projective module is a split exact sequence.
3.5.5. Corollary. A free module is projective. Over a eld every module is projec-
tive.
3.5.6. Example. Let I R be an ideal. If R/I is projective, then there is a ring
decomposition R/I R

R.
56 3. EXACT SEQUENCES OF MODULES
3.5.7. Proposition. A direct summand in a projective module is projective.
Proof. Let F F

be projective and g : M F surjective. By 3.5.3 there is a


section v

: F F

M F

to (g, 1
F
). Then v(y) = p
M
v

(y, 0) is a section
to g and F is projective.
3.5.8. Proposition. A module is projective if and only if it is a direct summand in
a free module.
Proof. By 2.4.12 any module is a factor module of a free module. By 3.5.3 a
projective factor module has a section, and is therefore by 3.1.10 a direct summand.
3.5.9. Proposition. Let F

be a family of projective modules, then the direct sum

is a projective module.
Proof. Let N L be surjective. Then by 2.5.8
Hom
R
(

, N) Hom
R
(

, L)
is the product

Hom
R
(F

, N)

Hom
R
(F

, L)
which is surjective by 3.1.6. So

is projective.
3.5.10. Proposition. Let F, F

be projective modules. Then F


R
F

is projective.
Proof. F
R
F

is clearly a direct summand in a free module.


3.5.11. Proposition. Let R S be a ring homomorphism and F a projective
module. The change of ring module F
R
S is a projective S-module.
Proof. A direct summand of a free R-module is changed to a direct summand of a
free S-module.
3.5.12. Proposition. Any module M admits an exact sequence
F M 0
where F is a projective module. That is, any module is a factor module of a pro-
jective module.
Proof. Take F free, 2.4.12.
3.5.13. Exercise. (1) Let R = R
1
R
2
. Show that R
1
, R
2
are projective ideals in R.
(2) Show that the ideal (2)/(6) in the ring Z/(6) is projective.
(3) Show that the ideal (2)/(4) in the ring Z/(4) is not projective.
3.6. Injective modules
3.6.1. Denition. An R-module E is an injective module if for any injective ho-
momorphism M N the homomorphism
Hom
R
(N, E) Hom
R
(M, E)
is surjective.
3.6.2. Proposition. Let E be an R-module. The following conditions are equiva-
lent:
(1) E is injective.
3.6. INJECTIVE MODULES 57
(2) The contravariant functor Hom
R
(, E) is exact.
(3) Given an exact sequence
M

N

L
then the following sequence is exact
Hom
R
(L, E)

Hom
R
(N, E)

Hom
R
(M, E)
(4) Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is short exact
0

Hom
R
(L, E)

Hom
R
(N, E)

Hom
R
(M, E)

0
Proof. This is clear from 3.3.2. (1) (2): This follows since the contravariant
functor Hom
R
(, E) is allways right exact. (1) (3) (4): This is true for any
exact contravariant functor.
3.6.3. Proposition. A module E is an injective module if and only if any injective
homomorphism 0 E L has a retraction.
Proof. Assume E injective and f : E L injective. Then Hom
R
(L, E)
Hom
R
(E, E) 0 is exact. So there exists a u : L E such that u f = 1
E
.
Then u is a retraction. Conversely given f : M N injective and h : M E.
Let L = Cok M E N, x h(x) f(x) and i
E
: E L, i
N
: N L the
injections, then i
N
f = i
E
h. Now i
E
is injective since f is. Let u : L E be
a retraction of i
E
, then h

= u i
N
satises h = h

f.
0

M
h

N
i
N

.
E
i
E

L
u

3.6.4. Corollary. A short exact sequence


0

E
f

N
g

L

0
where E is an injective module is a split exact sequence.
3.6.5. Example. Let I R be an ideal. If I is injective, then there is a ring
decomposition R/I R

R.
3.6.6. Proposition. A direct summand in an injective module is injective.
Proof. Let E E

be an injective module and f : E L an injective homomor-


phism. By 3.6.3 there is a retraction u

: L E

E E

to (f, 1
E
). Then
u(y) = p
E
u

(y, 0) is a retraction to f and E is injective.


3.6.7. Proposition. Let E

be a family of injective modules, then the direct product

is an injective module
58 3. EXACT SEQUENCES OF MODULES
Proof. Let M N be injective. Then by 2.5.8
Hom
R
(N,

) Hom
R
(M,

)
is the product

Hom
R
(N, E

Hom
R
(M, E

)
which is surjective by 3.1.6. So

is injective.
3.6.8. Proposition. A module E is injective, if for any ideal I R
Hom
R
(R, E) Hom
R
(I, E) 0
is exact.
Proof. Let f : E L be an injective homomorphism. The set of submodules
f(E) L

L and retractions u

: L

E is nonempty and inductively ordered.


By Zorns lemma a maximal L

, u

exists. If L/L

,= 0 choose a y LL

. The
homomorphism Ann(y + L

) E, a u

(ay) extends to u

: R E by
hypothesis. The setting L

+Ry E, x +ay u

(x) +u

(a) is a well dened


retraction. This contradicts maximality.
3.6.9. Denition. Let R be a domain. M is a divisible module if scalar multipli-
cation with a nonzero a R is surjective.
3.6.10. Proposition. (1) An injective module over a domain is divisible.
(2) A divisible module over a principal ideal domain is injective.
(3) Over a eld any module is injective.
Proof. (1) Let E be an injective module and a ,= 0. Choose x E and look
at 1
x
: R E. Scalar multiplication a
R
is injective, so there is an extension
1
y
: R E such that 1
x
= 1
y
a
R
. Then ay = x. (2) Let E be divisible. To
extend f : (a) E, choose x E such that ax = f(a), then 1
x
: R E, 1 x
extends f. (3) This is clear.
3.6.11. Proposition. Let R S be a ring homomorphism and E an injective
R-module. The induced module Hom
R
(S, E) is an injective S-module.
Proof. Let M N be an injective homomorphism of S-modules. Then by
2.7.10 Hom
S
(N, Hom
R
(S, E)) Hom
S
(M, Hom
R
(S, E)) is Hom
R
(N, E)
Hom
R
(M, E) being surjective since E is an injective R-module.
3.6.12. Lemma. (1) Qand Q/Zare divisible and therefore injective Zmodules.
(2) The homomorphism 2.5.11
x ev
x
: M Hom
Z
(Hom
Z
(M, Q/Z), Q/Z)
is injective.
Proof. (1) Clear by 3.6.10. (2) Let x ,= 0 and choose h : Z/ Ann(x) Q/Z
nonzero. Now Z/ Ann(x) Zx M so extend h to h

: M Q/Z. Then
ev
x
(h

) = h(1) ,= 0.
3.6.13. Theorem. Any module M admits an exact sequence
0 M E
where E is an injective module. That is, any module is a submodule of an injective
module.
3.6. INJECTIVE MODULES 59
Proof. By 2.4.12 choose a surjection F Hom
Z
(M, Q/Z) where F is a free
R-module. Then
0 M Hom
Z
(F, Q/Z)
is exact by 3.6.12. Since F

R the module
Hom
Z
(F, Q/Z)

Hom
Z
(R, Q/Z)
is injective, 3.6.7 and 3.6.11.
3.6.14. Proposition. A homomorphism f : M N is injective if and only if
Hom
R
(N, E) Hom
R
(M, E) 0
is surjective for any injective module E.
Proof. Assume that Ker f ,= 0 and choose 0 Ker f E, 3.6.13. The sequence
Hom
R
(N, E) Hom
R
(M, E) Hom
R
(Ker f, E) 0
is exact. So Hom
R
(N, E) Hom
R
(M, E) is not surjective.
3.6.15. Lemma. Given an injective homomorphism 0

N
i

M
, then the fol-
lowing conditions are equivalent:
(1) Any nonzero submodule L M has nonzero retraction i
1
(L) ,= 0.
(2) Given a homomorphismf : M K such that the restriction f i is injective,
then f is injective.
3.6.16. Denition. An injective homomorphism 0

N
i

M
satisfying the
equivalent conditions 3.6.15 is an essential extension. An essential extension
0

M
i

E
is an injective envelope of M.
3.6.17. Proposition. Any module M has an injective envelope. If M E, E

are
two injective envelopes, then there is an isomorphism f : E E

xing M.
Proof. By 3.6.13 choose M E

with E

injective. By Zorns lemma choose


M E E

maximal among the essential extensions of M. If E ,= E

then the
set of modules 0 ,= N

such that E N

is nonempty by maximality of E.
Let N be maximal among these by Zorns lemma. Since E

is injective there is a
commutative diagram
E

E

E

/N

The composite f : E

/N E

has E Imf. If Imf ,= E then by


maximality of E there is a submodule 0 ,= L Imf with E L = 0. Now
N f
1
(L) and E f
1
(L) = E L = 0 gives a contradiction. Therefore
Imf = E and f is a retraction making E a direct summand in E

. By 3.6.6 E is
an injective module. Now uniqueness: given two envelopes, let f : E E

be
any homomorphism xing M. Then f is injective, since M E is essential. If f
is not surjective, then E

Imf E

contradicting that M E

is essential.
3.6.18. Exercise. (1) Let R be a domain. Show that the fraction eld is an injective
module.
60 3. EXACT SEQUENCES OF MODULES
(2) Let R be a domain. Show that the torsion free divisible module injective.
(3) Show that for a ring that all modules are projective if and only all modules are injec-
tive.
3.7. Flat modules
3.7.1. Denition. An R-module F is a at module if for any injective homomor-
phism M N the homomorphism
M
R
F N
R
F
is injective.
3.7.2. Proposition. Let F be an R-module. The following conditions are equiva-
lent:
(1) F is a at.
(2) The functor
R
F is exact.
(3) Given an exact sequence of R-modules
M

N

L
Then the sequence of change of rings modules is exact
M
R
F

N
R
F

L
R
F
(4) Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is exact
0

M
R
F

N
R
F

L
R
F

0
Proof. This is follows from 3.3.2. (1) (2): This follows since
R
is allways
right exact. (1) (3) (4): This is true for any exact functor.
3.7.3. Proposition. A direct summand in a at module is at.
Proof. Let F F

be at and M N injective. Then M


R
(F F

)
N
R
(F F

) is injective. Conclusion by 2.6.11.


3.7.4. Proposition. Let F

be family of at modules, then the direct sum

is a at module.
Proof. Let M N be injective. Then M
R
(

) N
R
(

) is
injective by 2.6.11 and 3.1.6, so the product is at.
3.7.5. Corollary. A free module is at.
Proof. R is at since M
R
R M, so

R is at.
3.7.6. Corollary. A projective module is at.
Proof. By 3.5.8 a projective module is a direct summand in a free. Conclusion by
3.7.4.
3.7.7. Proposition. Let F, F

be at modules. Then F
R
F

is at.
Proof. Let M N be injective. Then by 2.6.10 M
R
(F
R
F

) N
R
(F
R
F

) is (M
R
F)
R
F

(N
R
F)
R
F

being injective by using the


denition twice.
3.7. FLAT MODULES 61
3.7.8. Proposition. Let R S be a ring homomorphism and F a at R-module.
The change of ring module F
R
S is a at S-module.
Proof. Let M N be an injective homomorphism of S-modules. Then by 2.7.5
M
S
(F
R
S) N
S
(F
R
S) is M
R
F N
R
F being injective since
F is a at R-module.
3.7.9. Theorem. Let R be a ring and F a module. The following conditions are :
(1) F is a at module.
(2) Hom
R
(F, E) is an injective module for any injective module E.
Proof. Let M N be injective. By 3.6.14 M
R
F N
R
F is injective if and
only if Hom
R
(N
R
F, E) Hom
R
(M
R
F, E) is surjective for any injective
module E. By 2.6.13 this is Hom
R
(N, Hom
R
(F, E)) Hom
R
(M, Hom
R
(F, E)).
3.7.10. Corollary. Given a short exact sequence
0

M

N

F

0
where F is a at module. Then M is at if and only if N is at.
Proof. Let E be an injective module. By 3.7.9 and 3.6.4 the sequence
0

Hom
R
(F, E)

Hom
R
(N, E)

Hom
R
(M, E)

0
is split exact. By 3.6.6 and 3.6.7 Hom
R
(N, E) is injective if and only if Hom
R
(M, E)
is so. Conclusion by 3.7.9.
3.7.11. Corollary. Given a short exact sequence
0

M

N

F

0
where F is a at module. For any module L there is a short exact sequence
0

L
R
M

L
R
N

L
R
F

0
Proof. Let E be an injective module. By 3.7.9 and 3.6.4 the sequence
0

Hom
R
(F, E)

Hom
R
(N, E)

Hom
R
(M, E)

0
is split exact. So also the sequence
0

Hom
R
(L, Hom
R
(F, E))

Hom
R
(L, Hom
R
(N, E))

Hom
R
(L, Hom
R
(M, E))

0
is split exact. By 2.6.13 this is natural isomorphic to the sequence
0

Hom
R
(L
R
F, E)

Hom
R
(L
R
N, E)

Hom
R
(L
R
M, E)

0
Conclusion by 3.6.14.
62 3. EXACT SEQUENCES OF MODULES
3.7.12. Theorem. A module F is at, if for any ideal I R
0 I
R
F R
R
F
is exact.
Proof. By 3.7.9 it sufces to see that Hom
R
(F, E) is injective for any injective E.
By 3.6.8 this amounts to Hom
R
(R, Hom
R
(F, E)) Hom
R
(I, Hom
R
(F, E)) be-
ing surjective. By 2.6.13 this homomorphismis Hom
R
(R
R
F, E) Hom
R
(I
R
F, E) which is surjective since E is injective.
3.7.13. Corollary. A module F is at, if for any nite ideal J R
0 J
R
F R
R
F
is exact.
Proof. Let I be any ideal. Given

b
i
x
i
I
R
F such that

b
i
x
i
= 0, then

b
i
x
i
= 0 (b
i
)
R
M and therefore also in I
R
M.
3.7.14. Exercise. (1) Show that if a R is a nonzero divisor and M is a at modules,
then a
M
is injective and a is a nonzero divisor on M.
(2) Show that Z/(n) is not a at Z-module for n ,= 0, 1.
(3) Show that Z/(2) is a at Z/(6)-module.
(4) Show that Z/(2) is not a at Z/(4)-module.
4
Fraction constructions
4.1. Rings of fractions
4.1.1. Lemma. Let R be a ring and U R such that 1 U and for u, v U the
product uv U. On U R is dened a relation
(u, a) (u

, a

) there is v U such that vu

a = vua

(1) The relation is an equivalence relation.


(2) If (u, a) (u

, a

) and (v, b) (v

, b

) then (uv, va + ub) (u

, v

+
u

).
(3) If (u, a) (u

, a

) and (v, b) (v

, b

) then (uv, ab) (u

, a

).
Proof. The claims are proved by simple calculations. (1) Symmetry is clear. Re-
exive follows as 1 U. Transitive: if (u, a) (u

, a

), (u

, a

) (u

, a

) then
vu

a = vua

, v

= v

. Since multiplication is commutative (v

vu

)u

a =
v

vua

= (v

vu

)ua

give (u, a) (u

, a

). (2) From wu

a = wua

, w

b =
w

vb

follow that w

vv

wu

a = w

vv

wua

, wuu

b = wuu

vb

. So now
ww

(u

)(va +ub) = ww

(uv)(v

+u

) as needed. (3) This is similar.


4.1.2. Denition. Let R be a ring. U R is a multiplicative subset if 1 U
and for u, v U the product uv U. The ring of fractions U
1
R is given by
equivalence classes
a
u
= (u

, v

)[(u

, v

) (u, v)
on U R under the relation 4.1.1
(u, a) (u

, a

) there is v U such that vu

a = vua

The addition is
a
u
+
b
v
=
va +ub
uv
and the multiplication is
a
u

b
v
=
ab
uv
This is well dened and U
1
R is a ring, veried analog to the rational numbers Q
being a ring. The canonical ring homomorphism is
: R U
1
R, a
a
1
4.1.3. Theorem. Let : R S be a ring homomorphism and U R a multi-
plicative subset. If all elements in (U) S are units, then there exists a unique
63
64 4. FRACTION CONSTRUCTIONS
ring homomorphism

: U
1
R S such that =

.
R

S
U
1
R

Proof.

(
a
u
) = (a)(u)
1
is the well dened unique ring homomorphism. Ob-
serve that the elements of form bv
1
satisfy the rules for fractions.
4.1.4. Proposition. Let U R be a multiplicative subset.
(1)
a
1
,= 0 in U
1
R if and only if Ann(a) U = .
(2) The canonical ring homomorphism R U
1
R is injective if and only if U
consists only of nonzero divisors.
Proof.
a
1
= 0 if and only if va = 0 for some v U.
4.1.5. Corollary. If Ris a domain and U is the nonzero elements then K = U
1
R
is a eld and the canonical homomorphism identies R as a subring.
4.1.6. Denition. The eld K in 4.1.5 is the fraction eld of R.
4.1.7. Corollary. If R is a domain with fraction eld K and : R L is an
injective ring homomorphism into a eld, then there is a unique homomorphism
K L extending .
4.1.8. Denition. The total ring of fractions of R is U
1
R, where U is the set of
nonzero divisors of R.
4.1.9. Corollary. A ring is a subring of its total ring of fractions.
4.1.10. Proposition. Let U = u
n
the powers of an element u R. There is an
isomorphism
R[X]/(uX 1) U
1
R, X
1
u
The ring of fractions with only one denominator is of nite type.
Proof. A pair of inverse homomorphism are constructed by 1.6.7 and 4.1.3.
4.1.11. Denition. Let U = u
n
the powers of an element u R. The ring
U
1
R, 4.1.10, is denoted
R[
1
u
]
4.1.12. Proposition. If U V R are multiplicative, then there is a canonical
identication
V
1
R (V/1)
1
U
1
R,
a
v

a/1
v/1
4.1.13. Proposition. Let U R
1
R
2
be a multiplicative subset of a product of
rings and let U
i
R
i
be the projections of U. Then there is an isomorphism
U
1
(R
1
R
2
) U
1
1
R
1
U
1
2
R
2
(a
1
, a
2
)
(u
1
, u
2
)

_
a
1
u
1
,
a
2
u
2
_
4.2. MODULES OF FRACTIONS 65
Proof. Injective: If
a
1
u
1
= 0 and
a
2
u
2
= 0 then there is (v
1
, v
2
), (w
1
, w
2
) U
such that v
1
a
1
= 0 and w
2
a
2
= 0. It follows that (v
1
, v
2
)(w
1
, w
2
)(a
1
, a
2
) = 0.
Surjective: Given
_
a
1
u
1
,
a
2
u
2
_
then there is (u
1
, v
2
), (v
1
, u
2
) U. It follows that
(a
1
v
1
,a
2
v
2
)
(u
1
v
1
,u
2
v
2
)

_
a
1
u
1
,
a
2
u
2
_
.
4.1.14. Exercise. (1) Show that if U contains a nilpotent element, then U
1
R = 0.
(2) Show that
Ker R U
1
R = a R[ua = 0, for some u U
(3) Let U = u
1
, . . . , u
m
and u = u
1
u
m
. Then show that
U
1
R = u
n

1
R
(4) Let R
1
, R
2
be domains with fraction elds K
1
, K
2
. Show that the total ring of frac-
tions of R
1
R
2
is K
1
K
2
.
(5) Show that the ring

a
2
n
Q[a Z, n N
is of nite type over Z.
4.2. Modules of fractions
4.2.1. Lemma. Let R be a ring and U R multiplicative. On U M is dened
a relation
(u, x) (u

, x

) there is v U such that vu

x = vux

(1) The relation is an equivalence relation.


(2) If (u, x) (u

, x

) and (v, y) (v

, y

) then (uv, vx + uy) (u

, v

+
u

).
(3) If (u, a) (u

, a

) in U R 4.1.1 and (v, x) (v

, x

) then (uv, ax)


(u

, a

).
Proof. The claims are proved by simple calculations. See the proof 4.1.1.
4.2.2. Denition. Let R be a ring and U a multiplicative subset. The module of
fractions U
1
M is given by equivalence classes
x
u
on U M under the relation
4.2.1
(u, x) (u

, x

) there is v U such that vu

x = vux

The addition is
x
u
+
y
v
=
vx +uy
uv
and the U
1
R-scalar multiplication is
a
u

y
v
=
ay
uv
The canonical homomorphism is
M U
1
M, x
x
1
66 4. FRACTION CONSTRUCTIONS
4.2.3. Theorem. Let U R be a multiplicative subset. Given an R-module M, a
U
1
R-module N and a R-homomorphism f : M N, then there exists a unique
U
1
R-homomorphism f

: U
1
M N,
x
u

1
u
f(x) such that f = f

i.
M
i

N
U
1
M
f

Proof. The claims are proved by simple calculations. If (u, x) (u

, x

) then
vu

x = vux

and therefore vu

f(x) = vuf(x

) showing
1
u
f(x) =
1
u

f(x

). So the
map is well dened. The rest is similar.
4.2.4. Corollary. Let f : M N a homomorphism of R-modules. Then there is
a homomorphism U
1
f : U
1
M U
1
N,
x
u

f(x)
u
of U
1
R-modules.
4.2.5. Proposition. The construction 4.2.2
M U
1
M
and 4.2.4
f : M N U
1
f : U
1
M U
1
N
is a functor from R-modules to U
1
R-modules.
Proof. Follows from the denitions by simple calculations as in the proof of 4.2.3.
For example, U
1
(f + g) = U
1
f + U
1
g, follows from
f(x)+g(x)
u
=
f(x)
u
+
g(x)
u
.
4.2.6. Remark. The induced homomorphism relates to the canonical homomor-
phism such that the diagram is commutative.
M

U
1
M
U
1
f

U
1
N
That is, the canonical homomorphism is a natural homomorphism.
4.2.7. Proposition. Let U R be a multiplicative subset and M a module.
(1)
x
1
,= 0 in U
1
M if and only if Ann(x) U = .
(2) The canonical homomorphism M U
1
M is injective if and only if U
consists only of nonzero divisors on M.
Proof.
x
1
= 0 if and only if vx = 0 for some v U.
4.2.8. Proposition. If M

is a family of modules, then the homomorphism


U
1
(

U
1
M

is a natural isomorphism of U
1
R-modules.
4.3. EXACTNESS OF FRACTIONS 67
Proof. By 2.4.3 there is an R-homomorphism

U
1
M

which
gives the U
1
R-homomorphism in in question by 4.2.3. An inverse is given by
2.4.3. This is also the method of common denominators in a nite sum.

i
x
i
u
i
=
1

i
u
i

i
(
j=i
u
j
)x
i
4.2.9. Denition. Let U = u
n
the powers of an element u R. The R[
1
u
]-
module, 4.1.10, U
1
M is denoted
M[
1
u
]
4.2.10. Proposition. If U V R are multiplicative, then there is a canonical
identication
V
1
M (V/1)
1
U
1
M,
x
v

x/1
v/1
4.2.11. Exercise. (1) Show that if U contains a nilpotent element, then U
1
M = 0.
(2) Show that
Ker M U
1
M = x M[ux = 0, for some u U
(3) Let U = u
1
, . . . , u
m
and u = u
1
u
m
. Then show that
U
1
M = u
n

1
M
(4) Show that U
1
M = 0 if and only if U Ann(x) ,= for all x M.
(5) Show that the fraction homomorphism of a composition is the composition of the
respective fraction homomorphisms.
(6) Let M be a free R-module. Show that U
1
M is a free U
1
R-module
(7) Show that the homomorphism
(

N
Z)[
1
2
]

N
Z[
1
2
]
is not surjective.
4.3. Exactness of fractions
4.3.1. Theorem. Let R be a ring and U a multiplicative subset. The functor
U
1
() is exact. Given an exact sequence of R-modules
M
f

N
g

L
Then the following sequence is exact
U
1
M

U
1
N

U
1
L
Proof. If
y
u
U
1
N maps to
g(y)
u
= 0 then there is v U such that 0 = vg(y) =
g(vy). Choose x M such that f(x) = vy. Then
x
vu
maps to
f(x)
vu
=
y
u
proving
exactness.
4.3.2. Corollary. Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is exact
0

U
1
M

U
1
N

U
1
L

0
68 4. FRACTION CONSTRUCTIONS
If the rst sequence is split exact, also the second sequence is split exact.
4.3.3. Corollary. For a homomorphism f : M N there are natural isomor-
phisms of U
1
R-modules.
(1) U
1
Ker f Ker U
1
f.
(2) U
1
Imf ImU
1
f.
(3) U
1
Cok f Cok U
1
f.
Proof. Represent the statements using short exact sequences. (1) The kernel is
determined by the exact sequence 0 Ker f M N, 3.1.4. (3) The cokernel
is determined by the exact sequence M N Cok f 0, 3.1.5. (2) The
image is determined by the exact sequence 0 Ker f M Imf 0, 2.3.5,
3.1.4.
4.3.4. Corollary. For submodules N, L M there are natural identications of
U
1
R-submodules and factor modules.
(1) U
1
(M/N) = U
1
M/U
1
N.
(2) U
1
(N +L) = U
1
N +U
1
L.
(3) U
1
(N L) = U
1
N U
1
L.
Proof. Represent the statements using short exact sequences. (1) 0 N M
M/N 0 is short exact giving 0 U
1
N U
1
M U
1
(M/N) 0. The
wanted isomorphism follows form 3.1.5. (2) N + L is the image of N L M
so conclude by 4.3.3. (3) N L is the kernel of N L M so conclude by
4.3.3.
4.3.5. Corollary. For ideals I, J R there are natural identications in U
1
R.
(1) U
1
(R/I) = U
1
R/U
1
I.
(2) U
1
(I +J) = U
1
I +U
1
J.
(3) U
1
(I J) = U
1
I U
1
J.
(4) U
1
(IJ) = U
1
IU
1
J.
Proof. (1) (2) (3) These are special cases of 4.3.4. (4) Both sides have the same
generators
ab
u
, a I, b J, u U.
4.3.6. Proposition. Let R U
1
R be the canonical homomorphism.
(1) For an ideal I R the extended ideal
IU
1
R = U
1
I
(2) For an ideal J U
1
R the extended contracted ideal
U
1
(J R) = J
(3) For an ideal I R the contracted extended ideal
I IU
1
R R
Proof. (1) This is clear. (2) U
1
(J R) J is true for any ring homomorphism
1.2.6. If
b
u
J then
b
1
J giving b J R and
1
u
b =
b
u
U
1
(J R). (3) This
is true for any ring homomorphism 1.2.6.
4.4. TENSOR MODULES OF FRACTIONS 69
4.3.7. Proposition. Let R U
1
R be the canonical homomorphism. For an
ideal I R the contracted extended ideal
I = IU
1
R R
if and only if each u U is a nonzero divisor on R/I.
Proof. Apply the snake lemma 3.2.4 to
0

I

R/I

0
0

0

U
1
R/U
1
I U
1
R/U
1
I

0
and get the exact sequence
0 I IU
1
R R Ker(R/I U
1
(R/I)) 0
Conclusion from 4.2.7.
4.3.8. Corollary. Let P R be a prime ideal. Then U
1
P U
1
R is either a
prime ideal or the whole ring.
Proof. By 4.1.5 and 4.3.5 U
1
R/U
1
P = U
1
(R/P) is either 0 or a domain.
4.3.9. Corollary. Let R be a principal ideal domain. Then U
1
R is a principal
ideal domain.
Proof. A restricted ideal is principal by hypothesis and the extension of a principal
ideal is principal. Conclude by 4.3.6.
4.3.10. Theorem. Let Rbe a unique factorization domain. Then U
1
Ris a unique
factorization domain.
Proof. By 4.3.7 the extension of an irreducible element is either a unit or an irre-
ducible element. Since a principal ideal (
a
u
) = (
a
1
), a factorization into irreducibles
in R gives a factorization in U
1
R. Now if (a) = (p
1
) . . . (p
n
) is a factorization
in R and (
a
1
) is irreducible in U
1
R. Then all but one
p
i
1
is a unit in U
1
R so
(
a
1
) = (
p
i
1
) is a prime ideal 4.3.9. The conditions 1.5.3 are satised.
4.3.11. Exercise. (1) Let U R be multiplicative. Show that
U
1
(R[X]) = (U
1
R)[X]
(2) Let : R S be a ring homomorphism and U R a multiplicative subset. Show
that
U
1
S = (U)
1
S
4.4. Tensor modules of fractions
4.4.1. Theorem. Let R be a ring and U a multiplicative subset. For any module
M, the homomorphism
M
R
U
1
R U
1
M, x
a
u

ax
u
is a natural isomorphism of U
1
R-modules.
Proof. By 2.6.3 there is an R-module homomorphism x
a
u

ax
u
. By denition
b
v
(x
a
u
) = x
ba
vu

bax
vu
=
b
v
ax
u
, so the this is a U
1
R-homomorphism. The
map U
1
M M
R
U
1
R,
x
u
x
1
u
is an inverse.
70 4. FRACTION CONSTRUCTIONS
4.4.2. Corollary. The two constructions, module change of ring to a fraction ring
and fraction module are natural isomorphic functors from modules to modules over
the fraction ring.
4.4.3. Corollary. Let R be a ring and U a multiplicative subset. Then U
1
R is a
at R-module.
Proof. This follows from 4.4.1 and 4.3.1.
4.4.4. Corollary. Let R be a ring and U a multiplicative subset and M, N mod-
ules. Then there is a natural isomorphism
U
1
(M
R
N) U
1
M
U
1
R
U
1
N
Proof. This follows from 2.7.5 and 4.4.1.
4.4.5. Corollary. Let I R be an ideal and U a multiplicative subset For a
module M module
U
1
(IM) U
1
IU
1
M
Proof. Use that IM = Im(I
R
M M) and 4.4.3,4.4.4.
4.4.6. Proposition. Let U R be a multiplicative subset and M an R-module.
Let i : M U
1
M be the canonical homomorphism and N U
1
M any
U
1
R-submodule. Then N is extended
U
1
(i
1
(N)) N
Proof. The inclusion U
1
(i
1
(N)) N is clear. If
x
u
N then
x
1
N giving
x i
1
(N) and
1
u
x =
x
u
U
1
(i
1
(N)).
4.4.7. Exercise. (1) Let F be a at R-module. Show that U
1
F is a at U
1
R-
module.
(2) Let M be a projective R-module. Show that U
1
M is a projective U
1
R-module.
(3) Show that the homomorphism
(

N
Z)
Z
Z[
1
2
]

N
Z
Z
Z[
1
2
]
is not surjective.
4.5. Homomorphism modules of fractions
4.5.1. Proposition. Let Rbe a ring and U a multiplicative subset. For any modules
M, N there is a natural homomorphism
U
1
Hom
R
(M, N) Hom
U
1
R
(U
1
M, U
1
N)
of U
1
R-modules.
Proof. Given f : M N and u U the setting
x
v

f(x)
uv
is a U
1
R-
homomorphism U
1
M U
1
N.
4.5.2. Proposition. Let R be a ring and U a multiplicative subset. For any R-
module M and any U
1
R-modules N there is a natural isomorphism
Hom
R
(M, N) Hom
U
1
R
(U
1
M, N)
of U
1
R-modules.
4.6. THE POLYNOMIAL RING IS FACTORIAL 71
Proof. This is the change of rings isomorphism 2.7.6 interpreted according to
4.4.2. Also this is a reinterpretation of 4.2.3.
4.5.3. Example. Let R be a ring and U a multiplicative subset. R U
1
R the
canonical homomorphism. The induced module functor maps an R-module M to
the U
1
R-module Hom
R
(U
1
R, M). The natural isomorphism 2.7.10 is
Hom
R
(N, M) Hom
U
1
R
(N, Hom
R
(U
1
R, M))
for any U
1
R-module N.
4.5.4. Example. The homomorphism 4.5.1 is in general neither injective nor sur-
jective.
(1) Not surjective:
0 = Hom
Z
(Q, Z) Hom
Q
(Q, Q) = Q
(2) Not injective:
0 ,= Hom
Z
(Q, Q/Z)
Z
Q Hom
Q
(Q, Q/Z
Z
Q) = 0
4.5.5. Exercise. (1) Let M be a Z-module and N a Q-module. Show that there is an
isomorphism Hom
Z
(M, N) Hom
Q
(M
Z
Q, N).
4.6. The polynomial ring is factorial
4.6.1. Denition. Let R be a unique factorization domain and let f = a
n
X
n
+
+ a
0
be a polynomial over R. Then the content of polynomial f, c(f), is the
greatest common divisor of the coefcients a
0
, . . . , a
n
.
4.6.2. Lemma (Gauss lemma). Let Rbe a unique factorization domain. For poly-
nomials f, g R[X]
c(fg) = c(f)c(g)
Proof. Assume by cancellation that c(f), c(g) are units in R. For any irreducible
p R the projections of f, g in R/(p)[X] are nonzero. Since R has unique factor-
ization the ideal (p) is a prime ideal. It follows that the projection of the product
fg in R/(p)[X] is also nonzero and therefore p is not a common divisor of the
coefcients of the product fg.
4.6.3. Theorem. Let R be a unique factorization domain. Then the ring of poly-
nomials R[X] is a unique factorization domain.
Proof. Let K be the fraction eld of R, then the polynomial ring K[X] is a prin-
cipal ideal domain. Let f R[X] and use unique factorization in K[X] to get
0 ,= a R and p
1
, . . . , p
n
R[X], irreducible in K[X], such that
af = p
1
. . . p
n
Assume by 4.6.2 that a = 1 and c(p
1
), . . . , c(p
n
) are units in R. Apply 4.6.2 and
1.6.5 to see that p
1
, . . . , p
n
are irreducible in R[X]. An irreducible p Rgenerates
a prime ideal (p) R[X]. A non constant irreducible p R[X] generates a prime
ideal (p) K[X] and therefore also a prime ideal (p) R[X]. So conditions
1.5.3 are satised.
4.6.4. Corollary. Let K be a eld. Then the polynomial ring K[X
1
, . . . , X
n
] is a
unique factorization domain.
Proof. Follows by induction from 4.6.3.
72 4. FRACTION CONSTRUCTIONS
4.6.5. Exercise. (1) Let f Z[X] be monic and assume f = gh where g, h Q[X]
are monic. Show that g, h Z[X].
(2) Let f Z[X] be monic and irreducible in Z/(n)[X]. Show that f is irreducible
Q[X].
(3) Let K be a eld. Show that the polynomial ring K[X
1
, X
2
, . . . ] in countable many
variables is a unique factorization domain
5
Localization
5.1. Prime ideals
5.1.1. Theorem (Krull). A nonzero ring contains a maximal ideal.
Proof. The nonempty set of ideals different from R is ordered by inclusion. Given
an increasing chain I

then I

is an ideal different from R which is a maximum


for the chain. Conclusion by Zorns lemma.
5.1.2. Corollary. Any proper ideal in a ring is contained in a maximal ideal.
Proof. If I ,= R then the factor ring R/I ,= 0 and contains a maximal ideal.
Conclusion by 1.2.10.
5.1.3. Theorem. Let P
1
, . . . , P
n
R be ideals with at most 2 not being prime
ideals. If an ideal
I P
1
P
n
then I P
i
for some i.
Proof. Assume n > 1 and I is not contained in any sub union. Moreover assume
the numbering such that P
3
, . . . , P
n
are prime ideals. Then for each i there is
a
i
(I P
i
)
j=i
P
j
The element
a
n
+a
1
. . . a
n1
is in I but not in any P
i
, giving a contradiction. So n = 1.
5.1.4. Proposition. Let P
1
, . . . , P
n
R be prime ideals and I any ideal. If for
some a R
a +I P
1
P
n
then I P
i
for some i.
Proof. If a
i
P
i
then conclusion by 5.1.3. On the contrary after renumbering
there exists j with 1 j < n such that
a P
1
P
j
P
j+1
P
n
Assume no inclusions between the prime ideals and choose by 5.1.3
b I P
j+1
P
n
P
1
P
j
Then a +b /
i
P
i
contradicts the hypothesis.
5.1.5. Theorem. Let R U
1
Rbe the canonical homomorphism. Extension and
contraction gives a bijective correspondence between prime ideals in R disjoint
from U and all prime ideals in U
1
R.
(1) For a prime ideal P RU the extended ideal PU
1
R is a prime ideal in
U
1
R and the contracted PU
1
R R = P.
73
74 5. LOCALIZATION
(2) For a prime ideal Q U
1
R the contracted ideal Q R is a prime ideal
and the extended (Q R)U
1
R = Q
R U
1
R
P PU
1
R
Q R Q
Proof. Conclusion by 4.3.6, 4.3.7, 4.3.8.
5.1.6. Corollary. Let R be a ring and U a multiplicative subset.
(1) An ideal P R is maximal among the ideals disjoint from U if and only if
PU
1
R is a maximal ideal in U
1
R.
(2) An ideal maximal among the ideals disjoint from U is a prime ideal.
(3) Any ideal disjoint from U is contained in a prime ideal disjoint form U.
Proof. The prime ideals disjoint from U are the prime ideals in U
1
R.
5.1.7. Theorem. The nilradical of a ring R is the intersection of all prime ideals
P.

0 =

P
P
Proof. By 1.3.8 the nilradical is contained in any prime ideal. Suppose u R is
not nilpotent. Then u
n

1
R is nonzero. Then contraction of a maximal ideal,
5.1.1, is a prime ideal in R not containing u.
5.1.8. Corollary. Let R be a ring.
(1) The radical of an ideal I is the intersection of all prime idealsP containing I

I =

IP
P
(2) For ideals I, J R,

I J =

J.
(3) If U is a multiplicative subset, then U
1

0 =

0 in U
1
R.
If R is reduced, then U
1
R is reduced.
Proof. (1) Use 5.1.7 on the factor ring R/I. (2) This follows from (1). (3) Clearly
U
1

0. If
a
u
is nilpotent, then va
n
= 0 for some v U. So (va)
n
= 0 and
a
u
=
va
vu
U
1

0.
5.1.9. Denition. A prime ideal minimal for inclusion among prime ideals is a
minimal prime ideal.
5.1.10. Theorem. Any prime ideal of Q R contains a minimal prime ideal
P Q.
Proof. The set of prime ideals in R is ordered by inclusion. Given a decreasing
chain P

then P

is a prime ideal. Conclusion by Zorns lemma.


5.1.11. Corollary. Let R S be a subring and P R a minimal prime ideal.
Then there is a minimal prime ideal Q S contracting to P = Q R.
5.1.12. Proposition. Let R be a domain. The following conditions are equivalent:
(1) R is a unique factorization domain.
(2) Any nonzero prime ideal contains a nonzero principal prime ideal.
5.2. LOCALIZATION OF RINGS 75
Proof. (1) (2): A prime ideal P ,= 0 contains an irreducible element generating
a prime ideal. (2) (1): Let U be the multiplicative subset generated by generators
of principal prime ideals. Given a R nonzero and not a unit. If a / U then by
5.1.6 there is a prime ideal (a) P such that P U = . Such a P contains no
principal prime ideal, so a U and 1.5.3 are satised.
5.1.13. Exercise. (1) Let K R be an innite subeld and I, P
1
, . . . , P
n
any ideals.
Show that if I P
1
P
n
then I P
i
for some i.
(2) Let P, P
1
, P
2
be proper ideals. Show that if P is a maximal ideal and P
n
P
1
P
2
then P = P
1
of P = P
2
.
5.2. Localization of rings
5.2.1. Denition. A ring R which contains precisely one maximal ideal P is a
local ring and denoted (R, P). The residue eld of R is R/P denoted by k(P).
A ring homomorphism : R S of local rings (R, P), (S, Q) is a local ring
homomorphism if (P) Q.
5.2.2. Proposition. Let R be a ring.
(1) If (R, P) is a local ring, then RP is the set of units of R.
(2) If the subset of non units in R is an ideal P, then (R, P) is a local ring.
Proof. (1) If u / P then by 5.1.1 (u) = R and u is a unit. (2) Any ideal I ,= R
contains only non units, so I P.
5.2.3. Proposition. A ring homomorphism : R S of local rings (R, P), (S, Q)
is a local ring homomorphismif the extended ideal PS Qor the contracted ideal
Q R = P. The residue homomorphism k(P) k(Q) is a eld extension.
Proof. The contraction Q R is a prime ideal containing P. The rest is clear.
5.2.4. Lemma. Let R be a ring and P a prime ideal. Then U = RP is a multi-
plicative subset. The ring of fractions U
1
R is a local ring. The maximal ideal is
the extended ideal PU
1
R. The residue eld is U
1
R/PU
1
R which is canoni-
cal isomorphic to the fraction eld of R/P.
5.2.5. Denition. Let R be a ring, P a prime ideal and U = RP. The localized
ring at P is the local ring R
P
= U
1
R. The residue eld is denoted k(P) =
R
P
/PR
P
.
R

R
P

R/P

k(P)
Note that
k(P) = R
P
/PR
P
= (R/P)
P
= (R/P)
(0)
5.2.6. Proposition. Given a ring homomorphism R S and a prime ideal Q
S. Then P = Q R is a prime ideal and there is a local ring homomorphism
76 5. LOCALIZATION
(R
P
, PR
P
) (S
Q
, QS
Q
) tting into the following commutative diagram.
S

S
Q

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
R
P

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S/Q

k(Q)
R/P

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
k(P)

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Proof. This is clear from the constructions.
5.2.7. Example. Let the ring be Z.
(1) The local ring at (0) is the fraction eld Q = Z
(0)
.
(2) The local ring Z
(p)
for a prime number p is identied with a subring of Q
Z
(p)
=
m
n
[p not dividing n
The residue eld F
p
= Z
(p)
/(p).
(3) Any nonzero ideal in Z
(p)
is principal of the form (p
n
) for some n.
5.2.8. Proposition. Let (R, P) be a local ring. One of the following conditions is
satised:
(1) The characteristic char(R) = 0. P Z = (0) and Q R is a subeld.
Q k(P) is a eld extension.
(2) The characteristic char(R) = 0. P Z = (p), p a prime number. Z
(p)
R
is a local subring. F
p
k(P) is a eld extension.
(3) The characteristic char(R) = p
n
, a power of a prime number. Z/(p
n
) R
is a local subring. F
p
k(P) is a eld extension.
Proof. (1) (2) are clear by 5.2.3 and 5.2.7. (3) If the characteristic is nonzero then
any prime ideal contracts Q Z = (p). So a prime number q ,= p gives a unit in
R. There is a local homomorphism Z
(p)
R, 4.1.3. The nontrivial kernel is (p
n
)
by 5.2.7.
5.2.9. Example. (1) A eld K is a local ring with maximal ideal (0). The power
series ring K[[X]] is a local ring with maximal ideal (X) and residue eld K.
(2) Let (R, P) be a local ring. The power series ring R[[X]] is a local ring with
maximal ideal (P, X) and residue eld k(P, X) = k(P).
5.2.10. Proposition. Let R be a domain.
(1) The local ring at (0) is the fraction eld K = R
(0)
.
(2) Any local ring R
P
is identied with a subring of the fraction eld K.
(3) The intersection
R =

P
R
P
, P a maximal ideal
5.2.11. Proposition. Let R S be a product of rings.
5.3. LOCALIZATION OF MODULES 77
(1) A prime ideal is of the form P S or R Q for uniquely determined prime
ideals P R or Q S.
(2) The local ring at P S is identied with R
P
through the projection RS
R.
(3) The local ring at RQis identied with S
Q
through the projection RS
S.
5.2.12. Theorem. Let P be a prime ideal and R R
P
the canonical homomor-
phism. Extension and contraction gives a bijective correspondence between prime
ideals in R contained in P and all prime ideals in R
P
.
(1) For a prime ideal Q P the extended ideal QR
P
is a prime ideal in R
P
and
the contracted QR
P
R = Q.
(2) For a prime ideal Q

R
P
the contracted ideal Q

R P is a prime ideal
and the extended (Q

R)R
P
= Q

R R
P
P PR
P
Q QR
P
Q

R Q

Proof. This is a special case of 5.1.5.


5.2.13. Proposition. Let U R be a multilicative subset and P U = a prime
ideal. Then PU
1
R is a prime ideal in U
1
R and canonically
R
P
= (U
1
R)
PU
1
R
Proof. This follows from 4.1.12 and reects the fraction rule
a
u
/
w
v
=
av
uw
.
5.2.14. Corollary. Let Q P R be prime ideals. Then QR
P
is a prime ideal
in R
P
and canonically
R
Q
= (R
P
)
QR
P
5.2.15. Denition. The intersection of all maximal ideals in a ring is the Jacobson
radical.
5.2.16. Remark. The Jacobson radical contains the nilradical. In a local ring the
Jacobson radical is the maximal ideal.
5.2.17. Exercise. (1) Show that a local ring is never a product of two nonzero rings.
(2) Show that a Ris in the Jacobson radical if and only if 1+ab is a unit for all b R.
(3) Let p be a prime number. Describe the prime ideals in the ring Z
(p)
.
(4) Let P be a prime ideal. Show that k(P) is the fraction eld of R/P.
(5) Let (R, P) be a local ring. Showthat (R[[X]], (P, X) is a local ring and the canonical
homomorphism R R[[X]] is a local homomorphism.
5.3. Localization of modules
5.3.1. Denition. Let R be a ring, P a prime ideal and U = RP. For a module
M, the localized module at P is the module M
P
= U
1
M over the local ring R
P
.
For a homomorphism f : M N the localized homomorphism is f
P
: M
P

N
P
and the residue homomorphism is f(P) : M
R
k(P) N
R
k(P).
78 5. LOCALIZATION
The constructions are functors, 2.7.4, 4.2.5. If f : M N is a homomorphism,
then the following diagram is commutative.
N

N
P

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
M
P

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
N/PN

N(P)
M/PM

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
M(P)

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5.3.2. Proposition. Let R be a ring, P a prime ideal and M a module.
(1) M
P
M
R
R
P
are natural isomorphic exact functors.
(2) M
P
/PR
P
M
P
M
R
k(P) are natural isomorphic functors.
Proof. See 4.4.1.
5.3.3. Corollary. For a homomorphism f : M N
(1) (Ker f)
P
Ker f
P
.
(2) (Imf)
P
Imf
P
.
(3) (Cok f)
P
Cok f
P
.
Proof. See 4.3.3.
5.3.4. Corollary. Let Rbe a ring and P a prime ideal. For submodules N, L M
(1) (M/N)
P
M
P
/N
P
.
(2) (N +L)
P
N
P
+L
P
.
(3) (N L)
P
N
P
L
P
.
Proof. See 4.3.4.
5.3.5. Proposition. Let R be a ring and P a prime ideal. If M

is a family of
modules, then the homomorphism
(

)
P

(M

)
P
is an isomorphism of R
P
-modules.
Proof. See 4.2.8.
5.3.6. Proposition. Let R be a ring, P a prime ideal.
(1) For an R module M and an R
P
-module N there is a natural isomorphism
M
R
R
P

R
P
N M
R
N
(2) For an R module M, L there is a natural isomorphism
(M
R
L)
P
M
P

R
P
L
P
(3) For an R module M, L there is a natural isomorphism
(M
R
L)(P) M(P)
k(P)
L(P)
Proof. See 4.4.4 and 2.7.4.
5.4. THE LOCAL-GLOBAL PRINCIPLE 79
5.3.7. Proposition. Let U R be a multilicative subset and P U = a prime
ideal. Then PU
1
R is a prime ideal in U
1
R and canonically for any R-module
M
M
P
= (U
1
M)
PU
1
R
Proof. This follows from 4.2.10 extending the set of denominators.
5.3.8. Corollary. Let Q P R be prime ideals and M an R-module Then
QR
P
is a prime ideal in R
P
and canonically
M
Q
= (M
P
)
QR
P
5.3.9. Denition. Let R be a ring. F is a locally free module is F
P
is a free
R
P
-module for all prime ideals P.
5.3.10. Proposition. Let R be a ring. F is a locally free module if F
Q
is a free
R
Q
-module for all maximal ideals Q.
Proof. A prime ideal P Q is contained in a maximal ideal. By 5.3.8 F
P

(F
Q
)
P
Q
is free.
5.3.11. Example. A free module is a locally free module.
5.3.12. Exercise. (1) Let P R be a prime ideal and R S a ring homomorphism.
Show that R
P
S
P
is a ring homomorphism.
(2) Let Q S be a prime ideal and R S a ring homomorphism. Show that R
QR

S
Q
is a local ring homomorphism.
(3) Let R = K L be a product of elds. Show that ideal K 0 is locally free but
not free.
5.4. The local-global principle
5.4.1. Theorem. Let R be a ring and M a module. The following conditions are
equivalent:
(1) M = 0.
(2) M
P
= 0 for all prime ideals P.
(3) M
P
= 0 for all maximal ideals P.
Proof. (1) (2) (3) is clear. (3) (1): Let 0 ,= x M be given. Then
Ann(x) P is contained in a maximal ideal, 5.1.2. Clearly 0 ,=
x
1
M
P
contradicts (3).
5.4.2. Corollary. Let R be a ring and f : M N a homomorphism. The follow-
ing conditions are equivalent:
(1) f is injective.
(2) f
P
is injective for all prime ideals P.
(3) f
P
is injective for all maximal ideals P.
Proof. Use 5.4.1 on Ker f.
5.4.3. Corollary. Let R be a ring and f : M N a homomorphism. The follow-
ing conditions are equivalent:
(1) f is surjective.
(2) f
P
is surjective for all prime ideals P.
(3) f
P
is surjective for all maximal ideals P.
80 5. LOCALIZATION
Proof. Use 5.4.1 on Cok f.
5.4.4. Corollary. Let R be a ring and f : M N a homomorphism. The follow-
ing conditions are equivalent:
(1) f is an isomorphism.
(2) f
P
is an isomorphism for all prime ideals P.
(3) f
P
is an isomorphism for all maximal ideals P.
5.4.5. Corollary. Let R be a ring and
M
f

N
g

L
a sequence of homomorphisms. The following conditions are equivalent:
(1) The sequence is exact.
(2) The sequence
M
P
f
P

N
P
g
P

L
P
is exact for all prime ideals P.
(3) The sequence
M
P
f
P

N
P
g
P

L
P
is exact for all maximal ideals P.
5.4.6. Corollary. Let R be a ring and
0

M
f

N
g

L

0
a sequence of homomorphisms. The following conditions are equivalent:
(1) The sequence is short exact.
(2) The sequence
0

M
P
f
P

N
P
g
P

L
P

0
is short exact for all prime ideals P.
(3) The sequence
0

M
P
f
P

N
P
g
P

L
P

0
is short exact for all maximal ideals P.
5.4.7. Corollary. Let R be a ring and F a module. The following conditions are
equivalent:
(1) F is at.
(2) F
P
is at for all prime ideals P.
(3) F
P
is at for all maximal ideals P.
Proof. Let 0 M N. Use 5.3.6 and 5.4.2 on M
R
F N
R
F.
5.4.8. Proposition. Let R be a ring and M a module. Then there is an exact
sequence
0 M

P maximal
M
P
Proof. Let 0 ,= x M be given. Then Ann(x) P is contained in a maximal
ideal, 5.1.2. Clearly 0 ,=
x
1
M
P
.
5.5. FLAT RING HOMOMORPHISMS 81
5.4.9. Corollary. Let R be a ring. Then there is an injective ring homomorphism
R

P maximal
R
P
5.4.10. Corollary. Let R be a ring. The following conditions are equivalent:
(1) R is reduced.
(2) R
P
is reduced for all prime ideals P.
(3) R
P
is reduced for all maximal ideals P.
Proof. Use 5.1.8 and 5.4.9.
5.4.11. Exercise. (1) Let R be a ring and
0

M
f

N
g

L

0
a split exact sequence. Show that the localized sequence is split exact for all prime
ideals P.
5.5. Flat ring homomorphisms
5.5.1. Denition. A ring homomorphism R S is a at ring homomorphism if
S is a at R-module, 3.7.1.
5.5.2. Proposition. Let R S be a ring homomorphism. The following condi-
tions are equivalent:
(1) R S is a at ring homomorphism.
(2) If M N is injective, then the change of rings homomorphism M
R
S
N
R
S is injective.
(3) The change of rings functor
R
S is an exact functor.
(4) Given an exact sequence of R-modules
M

N

L
Then the sequence of change of rings modules is exact
M
R
S

N
R
S

L
R
S
(5) Given a short exact sequence
0

M
f

N
g

L

0
Then the following sequence is exact
0

M
R
S

N
R
S

L
R
S

0
Proof. This follows from the denition 3.7.1 and 3.7.2.
5.5.3. Corollary. (1) Let U R be a multiplicative subset. Then the canonical
ring homomorphism R U
1
R is at.
(2) Let P R be a prime ideal. Then the canonical ring homomorphism R
R
P
is at.
5.5.4. Corollary. Let R S be a at ring homomorphism. For a homomorphism
f : M N of R-modules there are natural isomorphisms of S-modules.
(1) Ker f
R
S Ker f 1.
(2) Imf
R
S Imf 1.
(3) Cok f
R
S Cok f 1.
82 5. LOCALIZATION
Proof. See 3.3.3. Represent the statements using short exact sequences. (1) The
kernel is determined by the exact sequence 0 Ker f M N, 3.1.4. (3) The
cokernel is determined by the exact sequence M N Cok f 0, 3.1.5. (2)
The image is determined by the exact sequence 0 Ker f M Imf 0,
2.3.5, 3.1.4.
5.5.5. Corollary. Let R S be a at ring homomorphism. For submodules
N, L M there are natural identications of S-submodules and factor modules.
(1) (M/N)
R
S = M
R
S/N
R
S.
(2) (N +L)
R
S = N
R
S +L
R
S.
(3) (N L)
R
S = N
R
S L
R
S.
Proof. See 3.3.4. Represent the statements using short exact sequences. (1) 0
N M M/N 0 is short exact giving 0 N
R
S M
R
S
(M/N)
R
S 0. The wanted isomorphism follows form 3.1.5. (2) N +L is the
image of N L M. (3) N L is the kernel of N L M.
5.5.6. Corollary. Let R S be a at ring homomorphism. For ideals I, J R
there are natural identications in S.
(1) I
R
S = IS.
(2) (I J)S = IS JS.
Proof. (1) The sequence 0 I
R
S R
R
S (R/I)
R
S 0. is exact.
(2) This follows from (1) and 5.5.4.
5.5.7. Proposition. Let : R S be a at ring homomorphism.
(1) If F is a at S-module, then the restriction of scalars makes F a at R-
module.
(2) If S T is a at ring homomorphism, then the composite R T is a at
ring homomorphism.
Proof. (1) There is a natural isomorphismMRF (M
R
S)
S
F and clearly
the composition of two exact functors is exact. (2) This follows from (1).
5.5.8. Proposition. Let : R S be a ring homomorphism. The following
conditions are equivalent:
(1) is at.
(2)
P
: R
P
S
P
is at for all prime ideals P R.
(3)
P
: R
P
S
P
is at for all maximal ideals P R.
Proof. Use 5.4.7.
5.5.9. Proposition. Let R S be a ring homomorphism. The following condi-
tions are equivalent:
(1) R S is at.
(2) R
P
S
Q
is at for all prime ideals Q S and P = Q R.
(3) R
P
S
Q
is at for all maximal ideals Q S and P = Q R.
Proof. R
P
S
Q
is the composite R
P
S
P
S
P QS
P
= S
Q
, 5.3.8. (1)
(2): The composite R
P
S
P
S
Q
is at by 5.5.8 and 5.5.3. (3) (1): If
0 M N is exact, then 0 M
R
R
P
N
R
R
P
is exact, 5.5.3. By the
hypothesis 0 M
R
R
P

R
P
S
Q
N
R
R
P

R
P
S
Q
is exact, so cancelation
and canonical isomorphism, 2.7.5 and 5.3.2, give 0 (M
R
S)
Q
(N
R
S)
Q
5.6. FAITHFULLY FLAT RING HOMOMORPHISMS 83
exact for all maximal ideals Q. By 5.4.2 0 M
R
S N
R
S exact and
R S is at.
5.5.10. Exercise. (1) Let R S and S T be at homomorphisms. Show that the
composite R S is at.
(2) Show that Qis a at but not faithfully at Z-module.
(3) Let R be a ring and I =

0 the nilradical. Show that IR[X] is the nilradical of
R[X].
5.6. Faithfully at ring homomorphisms
5.6.1. Denition. A at R-module F is faithfully at if for any R-module M ,= 0
the tensor product M
R
F ,= 0.
5.6.2. Proposition. (1) A nonzero free module is faithfully at.
(2) Let F, F

be faithfully at modules, then F


R
F

is faihfully at.
(3) Let R S be a ring homomorphism and F a faithfully at R-module. The
change of ring module F
R
S is a faithfully at S-module.
Proof. (1) This is clear from properties of tensor product. (2) F
R
F

is at by
3.7.7 and M ,= 0 gives M
R
F ,= 0 so M
R
F RF

,= 0. (3) F
R
S is at
by 3.5.8 and if N ,= 0 is an S-module (3) F
R
S is at by 3.5.8 and if N ,= 0 is
an S-module then N
S
(F
R
S) N
R
F ,= 0 by 2.7.5.
5.6.3. Proposition. Let F be an R-module. The following conditions are equiva-
lent:
(1) F is faithfully at.
(2) A homomorphism M N is injective if and only if M
R
F N
R
F is
injective.
(3) A sequence of R-modules
M
f

N
g

L
is exact if and only if the sequence
M
R
F

N
R
F

L
R
F
is exact.
(4) A sequence
0

M
f

N
g

L

0
is short exact if and only if the sequence
0

M
R
F

N
R
F

L
R
F

0
is short exact.
Proof. (1) (2): f : M N is injective if and only if Kerf = 0 if and only if
Ker f 1
F
= (Ker f)
R
F = 0. (1) (3) (4): M N L is exact if and
only if (Imf + ker g)/ Ker g = 0 = (Imf + ker g)/ Imf.
5.6.4. Proposition. Let F be a at R-module. The following conditions are equiv-
alent:
(1) F is faithfully at.
(2) PF ,= F for all maximal ideals P R.
(3) F k(P) ,= 0 for all maximal ideals P R.
84 5. LOCALIZATION
Proof. (1) (2) (3): k(P)
R
R F/PF ,= 0. (2) (1): For 0 ,= x M
let Ann(x) = I P be a maximal ideal. Then R/I
R
F R/P
R
F 0
and 0 R/I
R
F M
R
F 0 are exact, so M
R
F ,= 0.
5.6.5. Proposition. Let R be a ring and F a module. The following conditions are
equivalent:
(1) F is faithfully at.
(2) F
P
is faithfully at for all prime ideals P.
(3) F
P
is faithfully at for all maximal ideals P.
Proof. (1) (2) (3): 5.6.2. (3) (1): F is at by 5.4.7. If M ,= 0, then
M
P
,= 0 for some P. Therefore (M
R
F)
P
M
P

R
P
F
P
,= 0.
5.6.6. Denition. A ring homomorphism R S is a faithfully at ring homo-
morphism if S is a faithfully at R-module.
5.6.7. Proposition. Let R S be a ring homomorphism. The following condi-
tions are equivalent:
(1) R S is faithfully at.
(2) A homomorphism M N is injective if and only if the change of rings
homomorphism M
R
F N
R
S is injective.
(3) A sequence of R-modules
M
f

N
g

L
is exact if and only if the sequence
M
R
S

N
R
S

L
R
S
is exact.
(4) A sequence
0

M
f

N
g

L

0
is short exact if and only if the sequence
0

M
R
S

N
R
S

L
R
S

0
is short exact.
Proof. See 5.6.3.
5.6.8. Proposition. Let R S be a faithfully at ring homomorphism and F an
R-module.
(1) If F
R
S is a at S-module, then F is a at R-module.
(2) If F
R
S is a faithfully at S-module, then F is a faithfully at R-module.
Proof. (1) If 0 M N is exact, then 0 M
R
S
S
F
R
S N
R
S
S
F
R
S is exact. This is natural isomorphic to 0 M
R
F
R
S
N
R
F
R
S, so by 5.6.3 0 M
R
F N
R
F is exact. (2) If M ,= 0 then
M
R
S
S
F
R
S M
R
F
R
S ,= 0 and therefore M
R
F ,= 0.
5.6.9. Proposition. Let R S be a faithfully at ring homomorphism.
(1) If F is a faithfully at S-module, then the restriction of scalars makes F a
faithfully at R-module.
(2) If S T is a faithfully at ring homomorphism, then the composite R T
is a faithfully at ring homomorphism.
5.6. FAITHFULLY FLAT RING HOMOMORPHISMS 85
Proof. (1) F is a at R-module by 5.5.7. If M ,= 0 is an R-module, then M
R
F M
R
S
S
F ,= 0. (2) This is a special case of (1).
5.6.10. Proposition. Let R S be a ring homomorphism.
(1) If F is a faithfully at S-module such that the restriction of scalars makes F
a at R-module. Then R S is a at ring homomorphism.
(2) If F is a faithfully at S-module such that the restriction of scalars makes F a
faithfully at R-module. Then R S is a faithfully at ring homomorphism.
(3) If S T is a faithfully at ring homomorphism and the composite R T
is a faithfully at ring homomorphism. Then R S is a faithfully at ring
homomorphism.
(4) If S T is a faithfully at ring homomorphism and the composite R T
is a at ring homomorphism. Then R S is a at ring homomorphism.
Proof. Let M N be an R-homomorphism. Then M
R
S
S
F N
R
S
S
F
is natural isomorphic to M
R
F N
R
F. (1) Since F is at as R-module
and faithfully at as S-module, it follows by 5.6.3 that if M N is injective
then M
R
S N
R
S is injective. (2) Since F is faithfully at both as R-
module and S-module, it follows by 5.6.3 that M N is injective if and only if
M
R
S N
R
S is injective. (3),(4) These are special cases of (1),(2).
5.6.11. Proposition. Let R S be a faithfully at ring homomorphism. for any
ideal I R the extended contracted returns I, i.e.
I = IS R
Proof. Tensor the homomorphismR/I R/ISRwith S. The induced R/I
R
S R/IS R
R
S is canonically isomorphic to the identity S/IS S/IS.
By 5.6.3 R/I R/IS R is injective giving I = IS R.
5.6.12. Corollary. A faithfully at ring homomorphism R S is injective.
5.6.13. Proposition. A at local homomorphism (R, P) (S, Q) is faithfully
at.
Proof. PS Q ,= S so S is faithfully at by 5.6.4.
5.6.14. Theorem (going-down). Let : R S be a at ring homomorphism and
Q S a prime ideal. For any prime ideal P

P = QR there is a prime ideal


Q

Q contracting to P

= Q

R.
S
R
.
.
.
.
.
.
.
Q
P

Proof. The local homomorphism R


P
S
Q
is faithfully at. The ring k(P

)
R
S
Q
is nonzero and therefor contains a maximal ideal Q

. The contraction to S,
Q

= Q

S contracts to P

= R Q

.
86 5. LOCALIZATION
5.6.15. Proposition. Let R S be a at homomorphism. The following condi-
tions are equivalent:
(1) R S is faithfully at.
(2) Any prime ideal P R is the contraction P = Q R of a prime ideal
Q S.
S
R

Q
P
Proof. (1) (2): The ring S
P
/PS
P
k(P)
R
S ,= 0, so let Q

be a maximal
ideal. The contraction Q = Q

S is a prime ideal such that P = Q R. (2)


(1): Let M ,= 0 be an R-module. Then M
P
,= 0 for some P. Let P = QR, then
(M
R
S)
Q
M
P

R
P
S
Q
,= 0 as R
P
S
Q
is faithfully at.
5.6.16. Proposition. The inclusion R R[X
1
, . . . , X
n
] is a faithfully at homo-
morphism
Proof. The R-module R[X
1
, . . . , X
n
] is free.
5.6.17. Exercise. (1) Show that a free module is faithfully at.
(2) Let F be a faithfully at module and G a at module. Show that F G is faithfully
at.
(3) Show that Qis a at but not faithfully at Z-module.
(4) Let R be a ring and I =

0 the nilradical. Show that IR[X] is the nilradical of
R[X].
(5) Show that

R
P
over all maximal ideals P is a faithfully at R-module.
6
Finite modules
6.1. Finite modules
6.1.1. Denition. Let R be a ring. A nite module is generated by nitely many
elements. The nite free module with standard basis e
1
, . . . , e
n
is denoted R
n
.
6.1.2. Lemma. Let R be a ring and M a module. The following conditions are
equivalent:
(1) M is generated by n elements x
1
, . . . , x
n
.
(2) There is a surjective homomorphism R
n
M 0, e
i
x
i
.
Proof. See 2.4.12.
6.1.3. Proposition. Let R S be a ring homomorphism. If an R-module M is
generated by n elements x
1
, . . . , x
n
. Then the change of rings S-module M
R
S
is generated by x
1
1, . . . , x
n
1 over S.
Proof. Follows from 6.1.2 and 3.4.1
6.1.4. Corollary. Let R be a ring and U a multiplicative subset. If M is a nite
R-module, then U
1
M is a nite U
1
R-module.
6.1.5. Proposition. For a short exact sequence
0

M
f

N
g

L

0
the following hold
(1) If N is nite, then L is nite.
(2) If M, L are nite, then N is nite.
Proof. (1) If y
1
, . . . , y
n
generates N, then g(y
1
), . . . , g(y
n
) generates L. (2) Choose
u : R
n
M 0 and v : R
m
L 0 exact. By 3.5.5 there is w : R
m
N
such that g w = v. There is a diagram
0

R
n
u

R
n
R
m
fu+w

R
m
v

0
0

M
f

N
g

L

0
Conclusion by the snake lemma 3.2.4.
6.1.6. Corollary. Let
0

M
f

N
g

L

0
be a split exact sequence. Then N is nite if and only if M, L are nite.
Proof. Let u be a retraction of f. By 6.1.5 Imu = M is nite. The rest is contained
in 6.1.5.
87
88 6. FINITE MODULES
6.1.7. Corollary. Let f : M N be a homomorphism.
(1) If M is nite, then Imf is nite.
(2) If Ker f, Imf are nite, then M is nite.
(3) If N is nite, then Cok f is nite.
(4) If Imf, Cok f are nite, then N is nite.
Proof. Use 6.1.5 on the exact sequences 3.1.8.
6.1.8. Corollary. Let M, N be modules. Then M N is nite if and only if M
and N are nite.
Proof. Use 6.1.6.
6.1.9. Proposition. Let R be a ring and M, N nite modules. Then M
R
N is
nite.
Proof. Use 6.1.2 and 6.1.5. Let R
m
M and R
n
N be surjective. Then
R
m

R
R
n
M
R
N is surjective, 3.4.1.
6.1.10. Proposition. Let R be a ring and F a module. The following conditions
are equivalent:
(1) F is nite and projective.
(2) F is a direct summand in a nite free module.
Proof. Use 6.1.2 and 6.1.8.
6.1.11. Proposition. Let R be a ring and F a nite projective module.
(1) If N is nite, then Hom
R
(F, N) is nite.
(2) If N is projective, then Hom
R
(F, N) is projective.
Proof. Use 6.1.8.
6.1.12. Proposition. Let F be a nite projective module and E an injective mod-
ule.
(1) F
R
E is injective.
(2) Hom
R
(F, E) is injective.
Proof. (1) (2) Both modules become summands in injective modules.
6.1.13. Proposition. Let R be a ring and U a multiplicative subset. For a nite
module M the following hold:
(1) U
1
M = 0 if and only if there is a u U such that uM = 0.
(2) Ann(U
1
M) = U
1
Ann(M) = Ann(M)U
1
R in the ring U
1
R.
Proof. Let x
1
, . . . , x
n
generate M. (1) U
1
M = 0 if and only if u
1
x
1
= =
u
n
x
n
= 0. Put u = u
1
. . . u
n
. (2) Ann(M) = Ann(x
1
) Ann(x
n
). The exact
sequence 0 Ann(x
1
) R Rx
1
0 localizes to 0 U
1
Ann(x
1
)
U
1
R U
1
Rx
1
0 giving that U
1
Ann(x
1
) = Ann(U
1
Rx
1
). Conclude
by 4.3.4.
6.1.14. Proposition. Let R be a ring and M

a family of modules. For any nite


module N there is a natural isomorphism
Hom
R
(N,

Hom
R
(N, M

)
6.2. FREE MODULES 89
Proof. A homomorphism f : N

has an image in a nite sum.


6.1.15. Exercise. (1) Show that

n
Z/(n) is not a nite Z-module.
(2) Let K be a eld and R = K[X
1
, X
2
, . . . ] the polynomial ring in countable many
variables. Show that R is a nite module, but the ideal (X
1
, X
2
, . . . ) is not a nite
module.
6.2. Free modules
6.2.1. Denition. Let R be a ring and let R
n
be the free module with standard
basis e
1
, . . . , e
n
.
(1) Let A = (a
ij
) be a m n-matrix with m rows and n columns, where the
entry a
ij
R. Identify R
n
with n-columns. Then matrix multiplication
x = (x
j
) y = Ax, y
i
=

j
a
ij
x
j
denes a homomorphism R
n
R
m
.
(2) Let f : R
n
R
m
be a homomorphism. Then dene a m n-matrix A =
(a
ij
) by
f(e
j
) =

i
a
ij
e
i
6.2.2. Proposition. (1) The dictionary dened in 6.2.1 gives a canonical isomor-
phism between the module of mn-matrices and Hom
R
(R
n
, R
m
).
(2) Matrix multiplication corresponds to composition of homomorphisms and the
identity matrix corresponds to the identity homomorphism.
(3) Invertible matrices correspond to isomorphisms.
Proof. Do linear algebra homework.
6.2.3. Denition. Let R be a ring.
(1) Let A = (a
ij
) be a m n-matrix. The (m 1) (n 1) matrix derived
from A be deleting i-row and j-column is A
ij
.
(2) For a square matrix A the determinant is dened by row expansion and in-
duction:
det(a
11
) = a
11
det A =

j
(1)
1+j
a
1j
det A
1j
(3) The determinant of a k k-matrix derived from A by choosing entries from
k rows and columns is a k-minor of A.
(4) If A is a n n-matrix, then the cofactor matrix A

= (a

ij
) has entries
a

ij
= (1)
i+j
det A
ji
given by (n 1)-minors.
6.2.4. Proposition. (1) For a xed a R, the scaled determinant a det is the
unique map from n n-matrices to R, which satises
(a) multilinear: For n1 xed columns a
1
, . . . , a
j1
, a
j+1
, . . . , a
n
, the
map
R
n
R, x a det(a
1
, . . . , a
j1
, x, a
j+1
, . . . , a
n
)
90 6. FINITE MODULES
is a homomorphism:
a det(..bx +cy..) = ba det(..x..) +ca det(..y..)
(b) alternating: For two equal columns
a det(..x..x..) = 0
(c) normed:
a det(1
n
) = a
(2) The determinant is calculated by any expansion
det A =

i
(1)
i+j
a
ij
det A
ij
=

j
(1)
i+j
a
ij
det A
ij
(3) If A, B are n n-matrices then the product rule holds
det AB = det Adet B
(4) Let A be an nn-matrix with cofactor matrix A

. Then the matrix Cramers


rule holds
AA

= A

A = det A(1
n
)
Where (1
n
) is the n n identity matrix.
(5) A square matrix A is invertible if and only if det A is a unit in R.
Proof. More linear algebra homework. (1) First prove that any map satisfying (a)-
(c) must satisfy the rule for interchanging columns and the rule for a modication
of a column. Then continue proof by induction. (2) Use the uniqueness from (1).
(3) To get the product rule det AB = det Adet B, you x A. Then both sides
satisfy (1) with norm det A. (4) Writeout (2) and use alternating from (1).
6.2.5. Theorem. Let f : R
n
R
n
be a homomorphism represented by an n n-
matrix A.
(1) f is a surjective if and only if det A is a unit.
(2) f is injective if and only if det A is a nonzero divisor.
Proof. (1) If f is surjective, then a section is represented by a matrix B such that
AB = (1
n
). Then det A is a unit. Conversely by 6.2.4. (2) If det A is a nonzero
divisor, then by 6.2.4 A

A = det A(1
n
) gives an injective homomorphism, so f is
injective. If det A is a zero divisor, then there is number k < n such that Ann(1
minors) = = Ann(k minors) = 0 and 0 ,= b Ann((k + 1) minors).
Assume that the kminor from rst k rows and columns c
k+1
has c
k+1
b ,= 0. Let
c
j
(1)
k+1+j
be the k-minor from rst k rows and rst k + 1 columns excluding
number j and put c
j
= 0, j > k + 1. Then A(c
j
) is a column with entries being
(k + 1) minors so f((bc
j
)) = A(bc
j
) = bA(c
j
) = 0 and f is not injective.
6.2.6. Proposition. Let R be a ring and f : R
n
R
m
homomorphism.
(1) If f is surjective, then n m.
(2) If f is injective, then n m.
(3) f is an isomorphism if and only if m = n and f is surjective.
Proof. (1) If n < m let p : R
m
R
n
be the projection onto rst n coordinates.
Then fp is surjective and represented by an mm-matrix Awith m-column zero.
A section to f p is represented by an mm-matrix B such that BA = (1
m
). But
the product AB must have a zero m-column, so the contradiction gives n m.
(2) If n > m let i : R
m
R
n
be injection into rst m coordinates. Then i f is
6.3. CAYLEY-HAMILTONS THEOREM 91
injective and represented by an n n-matrix A with n-row zero. Then det A = 0
contradicting 6.2.5. So n m. (3) If f is an isomorphism then m = n by (1), (2).
If f is surjective, then f is an isomorphism by 6.2.5 (1) and 6.2.4 (5).
6.2.7. Theorem. (1) A nite free module has a nite basis.
(2) The number of elements in a basis for a nite free module is independent of
the basis.
Proof. Let F be nite free generated by n elements. If y
1
, . . . , y
m
is part of a
basis, then by projection F R
m
there is a surjective homomorphismR
n
R
m
.
Conclusion by 6.2.6.
6.2.8. Denition. The number of elements 6.2.7 in a basis for a nite free module
F is the rank, rank
R
F.
6.2.9. Proposition. If x
1
, . . . , x
n
generates a free module F of rank n, then they
constitutes a basis.
Proof. Choose a basis and an isomorphism f : R
n
F. The homomorphism
g : R
n
F, e
i
x
i
is surjective. The composite f
1
g : R
n
R
n
is
surjective and therefore by 6.2.6 an isomorphism. Then g is an isomorphism and
x
i
a basis.
6.2.10. Proposition. Let F, F

be nite free modules. Then


(1) F F

is free and rank


R
F F

= rank
R
F + rank
R
F

.
(2) F
R
F

is free and rank


R
F
R
F

= rank
R
F rank
R
F

.
(3) Hom
R
(F, F

) is free and rank


R
Hom
R
(F, F

) = rank
R
F rank
R
F

.
6.2.11. Exercise. (1) Let R
n
R
n
be a surjective homomorphism. Show that it is an
isomorphism.
6.3. Cayley-Hamiltons theorem
6.3.1. Remark. Let R be a ring and f : M M a homomorphism. By 2.1.13
view M as an R[X]-module, where Xx = f(x) for x M. The homomorphism
1.6.7, 2.6.9, R[X] Hom
R
(M, M), a a
M
, X f is a ring homomorphism.
The image is R[f] the smallest subring containing 1
M
, f. M is naturally a R[f]-
module and the R[X]-module above is the restriction of scalars.
6.3.2. Proposition. Let A be an n n-matrix and I the ideal generated by the
entries a
ij
. The polynomial
det(X1
n
A) = a
0
+a
1
X +. . . a
n1
X
n1
+X
n
has a
0
, . . . , a
n1
I and gives the relation
a
0
(1
n
) +a
1
A+. . . a
n1
A
n1
+A
n
= 0
as n n-matrix.
Proof. View R
n
as a module over the ring R[X], 6.3.1, with scalar multiplication
Xx = Ax, x R
n
Let p : R[X]
n
R
n
be the R[X]-homomorphism determined by p(e
j
) = e
j
and denote U = X(1
n
) A the n n-matrix with entries from R[X]. Then
p(Ue
j
) = p(Xe
j
Ae
j
) = 0 in R
n
, so p(Uy) = 0 for all y R[X]
n
. Let U
have cofactor matrix U

over R[X]. Then p(UU

e
j
) = 0. By Cramers rule 6.2.4,
92 6. FINITE MODULES
UU

e
j
= det U e
j
in R[X]
n
. Therefore p(det U e
j
) = det U p(e
j
) = det U e
j
=
0 in R
n
. By calculation
det U = a
0
+a
1
X + +a
n1
X
n1
+X
n
in R[X], with a
i
I and for all x R
n
det U x = (a
0
(1
n
) +a
1
A+. . . a
n1
A
n1
+A
n
)x = 0
6.3.3. Theorem. Let I R be an ideal and f : M M a homomorphism on
a nite module generated by n elements. Suppose Imf IM, then there exist
a
0
, . . . , a
n1
I such that
a
0
1
M
+a
1
f +. . . a
n1
f
n1
+f
n
= 0
in Hom
R
(M, M).
Proof. Let x
1
, . . . , x
n
generate M and write
f(x
j
) =

i
a
ij
x
i
for an n n-matrix A with entries a
ij
I. View R
n
, x Ax and M, x f(x)
as modules over the ring R[X], 6.3.1. Then the R-homomorphism p : R
n

M, e
j
x
j
is an R[X]-homomorphism. By 6.3.2
det(X1
n
A) = a
0
+a
1
X +. . . a
n1
X
n1
+X
n
gives 0 = p(det(X(1
n
) A) e
j
) = det(X(1
n
) A) x
j
. It follows that
(a
0
1
M
+a
1
f +. . . a
n1
f
n1
+f
n
)x
j
= 0
for all j.
6.3.4. Proposition. Let I R be an ideal and M a nite module. If IM = M
then I + Ann(M) = R. That is there is a I such that (1 +a)M = 0.
Proof. By 6.3.3
a
0
1
M
+ + 1
M
= (a
0
+ +a
n1
)1
M
+ 1
M
= 0
Put a = a
0
+ +a
n1
.
6.3.5. Corollary. Let I R be an ideal and M a nite module. If IM = M and
all elements 1 +I are nonzero divisors on M, then M = 0.
6.3.6. Corollary. Let I Rbe an ideal and N M a submodule. Suppose M/N
is a nite module and M = N +IM. Then I + (N : M) = R.
6.3.7. Proposition. Let R be a ring and M a nite module. If a homomorphism
f : M M is surjective, then it is an isomorphism.
Proof. Regard M, f as a module over R[X] 6.3.1. Then (X)M = M, so by 6.3.4
there is p R[X] such that 1 + pX Ann(M). For any u Ker f, calculate
u = u +p(f) f(u) = (1 +pX)u = 0. So f is an isomorphism.
6.3.8. Exercise. (1) Let

0M = M. Show that M = 0.
6.4. NAKAYAMAS LEMMA 93
6.4. Nakayamas lemma
6.4.1. Theorem (Nakayama). Let (R, P) be a local ring and M a nite module.
The following conditions are equivalent:
(1) M = 0.
(2) PM = M.
(3) M
R
k(P) = 0.
Proof. (1) (2) (3) is clear. (2) (1): Elements in 1 + P are units in R, so
by 6.3.5 M = 0.
6.4.2. Corollary. Let (R, P) be a local ring and N M a submodule. Suppose
M/N is a nite module and M = N +PM. Then N = M.
6.4.3. Corollary. Let (R, P) be a local ring and M, N nite modules. If M
R
N = 0, then M = 0 or N = 0.
Proof. If M, N ,= 0 then M
R
k(P), N
R
k(P) ,= 0 are vector spaces over
k(P). Now M
R
N
R
k(P) M
R
k(P)
k(P)
, N
R
k(P) ,= 0, giving the
statement.
6.4.4. Corollary. Let R be a ring and M a nite module. The following conditions
are equivalent:
(1) M = 0.
(2) PM
P
= M
P
for all prime ideals P.
(3) PM
P
= M
P
for all maximal ideals P.
(4) M
R
k(P) = 0 for all prime ideals P.
(5) M
R
k(P) = 0 for all maximal ideals P.
Proof. Combine 6.4.1 with 5.4.1.
6.4.5. Corollary. Let (R, P) be a local ring and f : M N a homomorphism.
Assume N is nite. The following conditions are equivalent:
(1) f is surjective.
(2) f(P) is surjective.
Proof. f is surjective if and only if Cok f = 0. Cok f is nite, so it is zero if and
only if Cok f
R
k(P) = Cok(f(P)) = 0, 6.4.1.
6.4.6. Corollary. Let R be a ring and f : M N a homomorphism. Assume N
is nite. The following conditions are equivalent:
(1) f is surjective.
(2) f(P) is surjective for all prime ideals P.
(3) f(P) is surjective for all maximal ideals P.
6.4.7. Corollary. Let (R, P) be a local ring and M a nite module. Let x
1
, . . . , x
n

M. The following conditions are equivalent:
(1) x
1
, . . . , x
n
generate M.
(2) x
1
1, . . . , x
n
1 generate M
R
k(P).
The minimal number of generators of M is rank
k(P)
M
R
k(P).
6.4.8. Corollary. Let R be a ring and M a nite module. Let x
1
, . . . , x
n
M.
The following conditions are equivalent:
(1) x
1
, . . . , x
n
generate M.
94 6. FINITE MODULES
(2) x
1
1, . . . , x
n
1 generate M
R
k(P) for all prime ideals P.
(3) x
1
1, . . . , x
n
1 generate M
R
k(P) for all maximal ideals P.
6.4.9. Proposition. Let f : R
n
R
m
be a homomorphism represented by an
mn-matrix A. The following statements are equivalent:
(1) f is surjective.
(2) n m and the ideal (mminors) = R.
Proof. (1) (2): If f is surjective, then for any maximal ideal f(P) is surjective
linear map. So n m and some m-minor is nonzero in k(P). Therefore (m
minors) is not contained in P, so (m minors) = R. (2) (1): f(P) is
surjective for all maximal ideals, so f surjective by 6.4.6.
6.4.10. Proposition. Let f : R
n
R
m
be a homomorphism represented by an
mn-matrix A. The following statements are equivalent:
(1) f is injective.
(2) n m and the ideal Ann(n minors) = 0.
Proof. (1) (2): See the proof 6.2.5 (2) and note that the construction only uses
n m. (2) (1): Assume f(x) = Ax = 0 and choose n rows in A to give a
n n-matrix B. Then Bx = 0 and by Cramers rule det Bx = 0. Since det B
run through all n-minors it follows that the coordinates of x belong to Ann(n
minors) = 0.
6.4.11. Exercise. (1) Let R be a domain and f : R
n
R
m
a homomorphism repre-
sented by an mn-matrix. Show that f is injective if and only if n m and some
n minor ,= 0.
6.5. Finite presented modules
6.5.1. Denition. Let R be a ring. A nite presented module is a module M
having an exact sequence
R
n
R
m
M 0
or equivalently there is a short exact sequence
0 N R
m
M 0
with N nite.
6.5.2. Proposition. (1) Let R S be a ring homomorphism and M a nite
presented R-module Then the change of rings S-module M
R
S is nite
presented.
(2) Let U R be a multiplicative subset and M a nite presented R-module,
then U
1
M is a nite presented U
1
R-module.
6.5.3. Proposition. For a short exact sequence
0

M
f

N
g

L

0
the following hold:
(1) If M, L are nite presented, then N is nite presented.
(2) If L is nite presented and N is nite , then M is nite.
(3) If N is nite presented and M is nite , then L is nite presented.
6.5. FINITE PRESENTED MODULES 95
Proof. (1) Choose u : R
n
M 0 and v : R
m
L 0 exact with nite
kernels. By 3.5.5 there is w : R
m
N such that g w = v. There is a diagram
0

R
n
u

R
n
R
m
fu+w

R
m
v

0
0

M
f

N
g

L

0
By the snake lemma 3.2.4 the sequence 0 Ker u Ker f u+w Ker v 0
is exact. By 6.1.5 Ker f u +w is nite. (2) Choose v : R
m
L 0 exact with
nite kernel and w : R
m
N such that g w = v. There is a diagram
0

0

R
m
w

R
m
v

0
0

M
f

N
g

L

0
By the snake lemma 3.2.4 the sequence 0 Ker w Ker v M Cok w
0 is exact. By 6.1.5 M is nite. (3) Choose w : R
m
N 0 exact with nite
kernel. Then v = g w : R
m
L 0 is exact and there is a diagram
0

0

R
m
w

R
m
v

0
0

M
f

N
g

L

0
By the snake lemma 3.2.4 the sequence 0 Ker w Ker v M 0 is exact.
By 6.1.5 Ker v is nite and L is nite presented.
6.5.4. Corollary. Let
0

M
f

N
g

L

0
be a split exact sequence. Then N is nite presented if and only if M, L are nite
presented.
Proof. By 3.1.13 there is a split exact sequence
0

L
v

N
u

M

0
so the statement follows from 6.5.3.
6.5.5. Corollary. Let f : M N be a homomorphism.
(1) If M is nite and Imf nite presented, then Ker f is nite.
(2) If Ker f, Imf are nite presented, then M is nite presented.
(3) If N is nite presented and Imf nite, then Cok f is nite presented.
(4) If Imf, Cok f are nite presented, then N is nite presented.
Proof. Use the sequences 3.1.8.
6.5.6. Proposition. Let R be a ring and M, N nite presented modules.
(1) M N is nite presented.
(2) M
R
N is nite presented.
96 6. FINITE MODULES
Proof. (1) This is clear from 6.5.4. (2) If M = R
n
then M
R
N is nite presented
by (1). In general chose u : R
n
M 0 exact with nite kernel. The sequence
Ker u
R
N R
n

R
N M
R
N 0 is exact. So Ker u 1
N
is nite.
Conclusion by 6.5.5.
6.5.7. Proposition. Given submodules N, L M. Then
(1) If M/N, M/L are nite and M/N +L is nite presented, then M/N L is
nite.
(2) If M/N, M/L are nite presented and M/N L is nite, then M/N +L is
nite presented.
Proof. Use the sequence 3.2.7
0

M/N L

M/N M/L

M/N +L

0
together with 6.5.3.
6.5.8. Theorem. Let M be a nite presented module and N

a family of modules.
Then there is a natural isomorphism
M
R

(M
R
N

)
Proof. If M = R this is an isomorphism. Then this is also an isomorphism for
M = R
n
since both functors respect nite direct sums. In general choose R
n

R
m
M 0 exact. There is a diagram
R
n

R
m

M
R

R
n

R
N

R
m

R
N

M
R
N


0
giving isomorphism by the ve lemma 3.2.8.
6.5.9. Theorem. Let R S be a at ring homomorphism.
(1) For a nite module M and any module N the natural homomorphism
Hom
R
(M, N)
R
S Hom
S
(M
R
S, N
R
S)
is injective.
(2) For a nite presented module M and any module N the homomorphism
Hom
R
(M, N)
R
S Hom
S
(M
R
S, N
R
S)
is a natural isomorphism.
Proof. (1) If M = R this is an isomorphism. Then this is also an isomorphism
for M = R
n
since both functors respect nite direct sums. In general choose
0 K R
n
M 0 exact. There is a diagram
0

Hom
R
(M, N)
R
S

Hom
R
(R
n
, N)
R
S

Hom
R
(K, N)
R
S

0

Hom
S
(M
R
S, N
R
S)

Hom
S
(R
n

R
S, N
R
S)

Hom
S
(K
R
S, N
R
S)
giving injectivity. (2) In (1) K is nite, so the last vertical map is injective. Con-
clusion by the ve lemma 3.2.8.
6.5.10. Corollary. Let R be a ring and U a multiplicative subset.
6.5. FINITE PRESENTED MODULES 97
(1) For a nite module M and any module N the natural homomorphism
U
1
Hom
R
(M, N) Hom
U
1
R
(U
1
M, U
1
N)
is injective.
(2) For a nite presented module M and any module N the homomorphism
U
1
Hom
R
(M, N) Hom
U
1
R
(U
1
M, U
1
N)
is a natural isomorphism.
Proof. Use that R U
1
R is at.
6.5.11. Corollary. Let P R be a prime ideal and M a nite presented module.
For any module N there is a natural isomorphism
Hom
R
(M, N)
P
Hom
R
P
(M
P
, N
P
)
6.5.12. Corollary. Let R be a ring and
M
f

N
g

L
a sequence of homomorphisms with L a nite presented module. The following
conditions are equivalent:
(1) The sequence is split exact.
(2) The sequence
M
P
f
P

N
P
g
P

L
P
is split exact for all prime ideals P.
(3) The sequence
M
P
f
P

N
P
g
P

L
P
is split exact for all maximal ideals P.
Proof. The statement for exactness is true in general 5.4.5. The sequence splits if
Hom
R
(L, N) hom
R
(N, N) is surjective. By 5.4.3 this holds if Hom
R
(L, N)
P

hom
R
(N, N)
P
is surjective for all maximal ideals. Now L is nite presented, so
the localized is Hom
R
P
(L
P
, N
P
) Hom
R
P
(N
P
, N
P
) giving the conclusion.
6.5.13. Theorem. Let R be a ring and F a module. The following conditions are
equivalent:
(1) F is at.
(2) For any module N and a relation 0 =

i
y
i
x
i
N
R
F, there exist
z
j
F and a
ij
R such that 0 =

i
a
ij
y
i
N and x
i
=

j
a
ij
z
j
F.
(3) For any relation 0 =

i
b
i
x
i
F, there exist z
j
F and a
ij
R such that
0 =

i
a
ij
b
i
R and x
i
=

j
a
ij
z
j
F.
Proof. (1) (2): Let f : R
n
N, e
i
y
j
, then 0 Ker f
R
F F
n

N
R
F is exact. By assumption (x
i
) Ker f
R
F, so (x
i
) =

j
a
ij
z
j
with (a
ij
) Ker f. (2) (3) is clear. (3) (1): Let I R be an ideal and

b
i
x
i
Ker(I
R
F F). Then 0 =

i
a
ij
b
i
and x
i
=

j
a
ij
z
j
. Now
calculate

b
i
x
i
=

i
a
ij
b
i
x
i
= 0. By 3.7.12 F is at.
6.5.14. Theorem. Let (R, P) be a local ring and F a nite presented module. The
following conditions are equivalent:
(1) F is free.
98 6. FINITE MODULES
(2) F is projective.
(3) F is at.
(4) P
R
F F is injective.
Proof. (1) (2) (3) (4) are clear. (4) (1): Choose x
i
F such that
x
i
1 give a basis for F
R
k(P). The homomorphism f : R
n
F, e
i
x
i
is surjective by 6.4.5 and 0 K R
n
F 0 is exact with K nite, 6.5.5.
There is a diagram
P
R
K

P
R
R
n

P
R
F

0
0

K

R
n

0
k(P)
R
K

k(P)
R
R
n


k(P)
R
F

0
By the snake lemma 3.2.4 k(P)
R
K = 0 and therefore K = 0 by 6.4.1, so F is
free.
6.5.15. Corollary. Let (R, P) be a local ring and f : F F

a homomorphism
of nite free modules. The following conditions are equivalent:
(1) f has a retraction u : F

F.
(2) f is injective and Cok f is free.
(3) f(P) is injective.
Proof. (1) (2): Cok f is projective, so free by 6.5.14. (2) (3) is clear. (3)
(1): Let f

: F

be the dual homomorphism. f

(P) is surjective, so f

is
surjective, 6.4.5. A section v of f

gives a retraction u = v

.
6.5.16. Corollary. Let R be a ring and F a nite presented module. The following
conditions are equivalent:
(1) F is projective.
(2) F is at.
(3) P
R
F F is injective for all maximal ideals P.
(4) F
P
is free for all maximal ideals P.
Proof. (1) (2) (3) are clear by 6.5.14. (3) (4): PR
P

R
P
F
P
F
P
is
(P
R
F)
P
F
P
and therefore injective. Conclusion by 6.5.14. (4) (1): Let
N L 0 be exact. By hypothesis Hom
R
P
(F
P
, N
P
) Hom
R
P
(F
P
, L
P
)
0 is exact for all maximal ideals. By 6.5.11 Hom
R
(F, N)
P
Hom
R
(F, L)
P
0
is exact for all maximal ideals. By 5.4.3 Hom
R
(F, N) Hom
R
(F, L) 0 is
exact and F is projective.
6.5.17. Exercise. (1) Show that a nite projective module is nite presented.
(2) Let I R be an ideal. Show that R/I is a nite presented R-module if and only if
I is a nite ideal.
(3) Show that Qis a at, but not projective Z-module.
6.6. FINITE RING HOMOMORPHISMS 99
6.6. Finite ring homomorphisms
6.6.1. Denition. A ring homomorphism : R S is a nite ring homomor-
phism if S is a nite R-module. If R S is a subring, then a nite ring homomor-
phism is a nite ring extension.
6.6.2. Proposition. Let R be a ring.
(1) Let f R[X] be a monic polynomial. Then the homomorphism R
R[X]/(f) is nite.
(2) Let f : M M be a homomorphism of a nite R-module. Then the homo-
morphism 6.3.1, R R[f] is nite.
Proof. (2) Follows from (1) and 6.3.3.
6.6.3. Lemma. Let : R S be a nite ring homomorphism.
(1) If N is a nite S-module, then by restriction along the R-module N is nite.
(2) Let U R be a multiplicative subset, then U
1
R U
1
S is a nite ring
homomorphism.
6.6.4. Proposition. Let R S be a nite ring extension of domains. Then R is a
eld if and only if S is a eld.
Proof. Let R be a eld, a minimal equation 6.3.3 for scalar multiplication by a
nonzero b S, b
S
as R-module homomorphism
b
n
+ +a
0
= 0
gives
b
1
= a
1
0
(a
n1
b
n2
+ +a
1
) S
Let S be a eld and 0 ,= a R. An equation 6.3.3 for scalar multiplication a
1
S
as
R-homomorphism
a
n
+ +a
0
= 0
gives
a
1
= (a
0
a
n1
+ +a
n1
) R
6.6.5. Corollary. Let R S be a nite ring homomorphism. A prime ideal Q S
is maximal if and only if the contraction Q R is maximal.
Proof. R/Q R S/Q is a nite extension of domains.
6.6.6. Theorem. Let R S be a nite ring extension and P R a prime ideal.
Then there is a prime ideal Q S contracting P = Q R.
S
R

Q
P
Proof. R
P
S
P
is a nite ring extension. Since S
P
,= 0 there is a maximal
ideal in S
P
contracting to PR
P
by 6.6.5. The corresponding prime ideal Q S
contracts to P.
100 6. FINITE MODULES
6.6.7. Corollary (going-up). Let R S be a nite ring homomorphism, Q S
a prime ideal and P = Q R the contraction. For any prime ideal P P

in R
there is a prime ideal Q Q

in S contracting P

= Q

R.
S
R
.
.
.
.
.
.
.
Q

Q
P

Proof. R/P S/Q is a nite ring extension. So there is a prime ideal in S/Q
contracting to P

/P, 6.6.6. The corresponding prime ideal Q Q

S contracts
to P

.
6.6.8. Theorem. Let R S be a nite ring extension and E an R-module. The
following conditions are equivalent:
(1) E is an injective R-module.
(2) Hom
R
(S, E) is an injective S-module.
Proof. (1) (2): This is 3.6.11. (2) (1): Let E E

be an injective envelope,
3.6.17. Hom
R
(S, E) Hom
R
(S, E

) is injective and identies Hom


R
(S, E) as
an essential submodule. For this, let f : S E

be nonzero and S = Rb
1
+
+ Rb
n
. Then use E E

essential to construct with renumbering a maximal


sequence a
1
, . . . , a
m
R such that
0 ,= a
1
f(b
1
) E, . . . , 0 ,= a
1
a
m
f(b
m
) E
Then a
1
a
m
f : S E is nonzero, so the extension of Homs is essential. Next,
by (1) (2) Hom
R
(S, E

) is an injective S module, so the extension of Homs


above is trivial, 3.6.17. It follows that
Hom
R
(S, E)

Hom
R
(S, E

Hom
R
(R, E)

Hom
R
(R, E

)
E

E

with the right down map being surjective. So E = E

and E is injective.
6.6.9. Exercise. (1) Show that Z Z[

5] is nite.
(2) Let p be a prime number. Show that Z Z
(p)
is not nite.
7
Modules of nite length
7.1. Simple modules
7.1.1. Denition. A nonzero module M is a simple module if 0 and M are the
only submodules.
7.1.2. Lemma. Let f : M M

be a homomorphism
(1) If M is simple, then f is either zero or injective.
(2) If M

is simple, then f is either zero or surjective.


(3) If M, M

are both simple, then f is either zero or an isomorphism.


Proof. This follows from 2.3.3. (1) Ker f M is either 0 or M. (2) Imf N is
either 0 or N.
7.1.3. Theorem. Let M be a simple R-module.
(1) M is isomorphic to the factor module R/P, where P = Ann(M) is a maxi-
mal ideal.
(2) M = Rx is nite and generated by any one nonzero element x M.
Proof. A nonzero f : R M must be surjective. If Ker f = P then and R/P
M is simple exactly when P is maximal. Clearly P = Ann(M).
7.1.4. Corollary. Any nonzero nite module has a simple factor module.
Proof. If M ,= 0 then M
P
,= 0 for some maximal ideal 5.4.1. By 6.4.1 M
R
k(P) = M/PM ,= 0. Choose a nonzero linear projection M/PM k(P),
giving M k(P) 0 exact.
7.1.5. Corollary. Let (R, P) be a local ring and M a nite module. The following
conditions are equivalent:
(1) M = 0.
(2) Hom
R
(M, k(P)) = 0.
7.1.6. Proposition. Let M, M

be simple modules. The following conditions are


equivalent:
(1) M M

.
(2) M
R
M

,= 0.
(3) Hom
R
(M, M

) ,= 0.
Moreover if M M

there are non natural isomorphisms


M
R
M

M Hom
R
(M, M

)
Proof. This follows from 7.1.3 with M = R/P.
7.1.7. Proposition. Let I R be an ideal. A simple R/I-module M is a simple
R-module such that I Ann(M).
101
102 7. MODULES OF FINITE LENGTH
Proof. Maximal ideals in R/I are of the form P/I and (R/I)/(P/I) R/P.
7.1.8. Proposition. Let U R be a multiplicative subset and M is a simple R-
module with Ann(M) = P.
(1) If U P = , then is U
1
M = M is a simple U
1
R-module.
(2) If U P ,= , then U
1
M = 0.
Proof. Let P R be a maximal ideal, then the fractions U
1
(R/P) is either 0 or
R/P, 5.1.5.
7.1.9. Proposition. Let R = R
1
R
2
be a product of rings. A simple R-module
is of the form M
1
0 or 0 M
2
, where M
i
is a simple R
i
-module.
Proof. Clear by 2.2.9.
7.1.10. Example. (1) If K is a eld, a one dimensional vector space is simple.
(2) If R is a principal ideal domain, then R/(p) is a simple module for all irre-
ducible (p).
(3) Z/(p) are simple for all prime numbers p.
(4) Let K be a eld. K[X]/(X a) is a simple K[X]-module.
7.1.11. Exercise. (1) Show that Z does not contain any simple modules.
(2) Show that Qdoes not have any simple factor Z-module.
(3) Let L, L

M be simple submodules. Show that either L L

= 0 of L = L

.
(4) Let L ,= L

M be simple submodules. Show that L L

= L +L

.
7.2. The length
7.2.1. Denition. Let R be a ring. A module M has nite length if it admits a
composition series by submodules
0 = M
0
M
1
M
n1
M
n
= M
such that each factor M
i
/M
i1
is a simple module.
7.2.2. Lemma. Any two nite composition series have the same number of sub-
modules.
Proof. Let l(M) be the least length of a composition series. M is simple if and
only if l(M) = 1. Let 0 = M
0
M
1
M
n
= M be any composition
series. For a submodule N M there is a ltration 0 = N M
0
N M
1

N M
n
= N with factors N M
i
/N M
i1
M
i
/M
i1
being either
simple or 0. It follows that l(N) l(M). If l(N) = l(M) then each factor is
nonzero, so N M
i
/N M
i1
= M
i
/M
i1
. This gives M
i
N and nally
N = M. Applying this to the composition series of M gives l(M
0
) < l(M
1
) <
< l(M
n
) so n l(M) as needed.
7.2.3. Denition. The number of nontrivial submodules in a ltration as above
will be denoted
R
(M) and is the length of M. Remark that a module of nite
length is simple if and only if the length is 1.
7.2.4. Theorem. Given an exact sequence
0

M
f

N
g

L

0
then the following statements are equivalent:
(1) N has nite length.
7.2. THE LENGTH 103
(2) M and L have nite length.
If N has nite length, then

R
(N) =
R
(M) +
R
(L)
Proof. (1) (2): A nite ltration N
i
with simple factors in N induces a ltration
f
1
(N
i
) in M and a ltration g(N
i
) in L with factors being simple or 0. (2)
(1): A nite composition series M
i
in M and L
j
in L may be spliced together
f(M) = g
1
(0) g
1
(L
1
) . . .
to give a composition series of N. The length formula follows from 7.2.2.
7.2.5. Corollary. Let f : M N be a homomorphism.
(1) M has nite length if and only if Ker f, Imf have nite length. If nite length

R
(M) =
R
(Ker f) +
R
(Imf)
(2) N has nite length if and only if Imf, Cok f have nite length. If nite length

R
(N) =
R
(Imf) +
R
(Cok f)
Proof. Use the sequences 3.1.8.
7.2.6. Corollary. Let f : M M be a homomorphism on a module of nite
length. Then

R
(Ker f) =
R
(Cok f)
and the following conditions are equivalent:
(1) f is injective.
(2) f is surjective.
(3) f is an isomorphism.
Proof. Use the sequences 3.1.8.
7.2.7. Corollary. Let N M be a submodule and suppose M has nite length.
(1)
R
(N)
R
(M).
(2)
R
(N) =
R
(M) if and only if N = M.
7.2.8. Corollary. Let N, L M be submodules and suppose N, L has nite
length. Then N +L, N L have nite length and

R
(N +L) +
R
(N L) =
R
(N) +
R
(L)
Proof. By 2.3.8 N/N L N +L/L.
7.2.9. Proposition. Given submodules N, L M. The following statements are
equivalent:
(1) M/N, M/L have nite length.
(2) M/N L has nite length.
If nite length

R
(M/N +L) +
R
(M/N L) =
R
(M/N) +
R
(M/L)
Proof. Use the exact sequence 3.2.7
0

M/N L

M/N M/L

M/N +L

0
together with 7.2.4.
104 7. MODULES OF FINITE LENGTH
7.2.10. Proposition. A R-module M of nite length is nite and generated by

R
(M) or less elements.
Proof. There is an exact sequence 0

N
f

M
g

L

0 with Lsimple, 7.1.4.
If x
i
generate N and g(y) generate L, then f(x
i
), y generate M. Conclusion by
induction.
7.2.11. Proposition. Let I R be an ideal. Suppose an R/I-module M has nite
length. Then M has nite length as R-module and

R/I
(M) =
R
(M)
Proof. From 7.1.7 follows that a composition series as R/I-module is also a com-
position series as R-module. and 7.2.2.
7.2.12. Proposition. Let U R be a multiplicative subset. Suppose an R-module
M has nite length. Then U
1
M has nite length as U
1
R-module and

U
1
R
(U
1
M)
R
(M)
Proof. This follows from 7.1.8 and 7.2.2. The fraction construction of a composi-
tion series may be rened to a composition series.
7.2.13. Proposition. Let (R, P) be a local ring and M a module of nite length.
The following conditions are equivalent:
(1) M = 0.
(2) Hom
R
(k(P), M) = 0.
(3) Hom
R
(M, k(P)) = 0.
Proof. Clear by a composition series as k(P) is the only simple module.
7.2.14. Proposition. Let M, M

be modules of nite length.


(1) M M

has nite length and


R
(M M

) =
R
(M) +
R
(M

).
(2) M
R
M

has nite length and


R
(M
R
M

)
R
(M)
R
(M

).
(3) Hom
R
(M, M

) has nite length and


R
(Hom
R
(M, M

))
R
(M)
R
(M

).
Proof. (1) This is clear from 7.2.4. (2) If M = R/P is simple then M
R
M

/PM

is a factor module, so of nite length and the equality follows from 7.2.4.
In general there is an exact sequence 0 N M L 0 with Lsimple giving
an exact sequence N
R
M

M
R
M

L
R
M

0. Conclusion by
induction and 7.2.4. (2) If M = R/P is simple then Hom
R
(M, M

) M

is a
submodule, so of nite length and the equality follows from 7.2.4. In general there
is an exact sequence 0 N M L 0 with L simple giving an exact
sequence 0 Hom
R
(L, M

) Hom
R
(M, M

) Hom
R
(N, M

). Conclusion
by induction and 7.2.4.
7.2.15. Proposition. Let R = R
1
R
2
be a product of rings. An R-module
M
1
M
2
is of the nite length if and only if M
i
is a nite length R
i
-module.
Moreover

R
(M
1
M
2
) =
R
1
(M
1
) +
R
2
(M
2
)
Proof. This follows from 7.1.9 and 7.2.11.
7.3. ARTINIAN MODULES 105
7.2.16. Example. Let K be a eld. A module M is of nite length if and only if it
is a nite dimensional vector space. Then

K
(M) = rank
K
(M)
7.2.17. Exercise. (1) Compute

Z
(Z/(p
n
1
1
. . . p
n
k
k
)) = n
1
+ +n
k
(2) Compute

K[X]
(K[X]/((X a
1
)
n
1
. . . (X a
k
)
n
k
)) = n
1
+ +n
k
7.3. Artinian modules
7.3.1. Lemma. Let M be a module. The following conditions are equivalent:
(1) Any descending chain M
i+1
M
i
of submodules is stationary: there
is n such that M
i+1
= M
i
for i > n.
(2) Any nonempty subset of submodules of M contains a minimal element.
Proof. (1) (2): Suppose a nonempty subset of submodules do not contain a
minimal element. Then choose a non stationary descending chain. (2) (1): A
descending chain containing a minimal element is stationary.
7.3.2. Denition. A module M which satises the conditions of 7.3.1 is an ar-
tinian module.
7.3.3. Example. (1) A module of nite length is artinian. Just note that a de-
scending chain produces a decreasing sequence of lengths.
(2) A vector space is artinian if and only if it is nite.
7.3.4. Theorem. Let
0

M
f

N
g

L

0
be an exact sequence of modules over the ring R. The following statements are
equivalent:
(1) N is artinian.
(2) M and L are artinian.
Proof. (1) (2): A chain M
i
in M gives a chain f(M
i
) in N which becomes
stationary. So M is artinian. A chain in L
i
in L gives a chain in g
1
(L
i
) in N,
which becomes stationary. Then also the original chain L
i
= g(g
1
(L
i
)) becomes
stationary and L is artinian. (2) (1): A chain N
i
in N, induces chains f
1
(N
i
)
in N and g(N
i
) in L which become stationary. By the snake lemma 3.2.4
f
1
(N
i
)/f
1
(N
i+1
) N
i
/N
i+1
g(N
i
)/g(N
i+1
)
is exact and the original chain is stationary.
7.3.5. Corollary. (1) Let N M be a submodule. Then M is artinian if and
only if both N and M/N are artinian.
(2) Let f : M N be a homomorphism.
(a) M is artinian if and only if Ker f, Imf are artinian.
(b) N is artinian if and only if Imf, Cok f are artinian
Proof. Use the sequences 3.1.8.
106 7. MODULES OF FINITE LENGTH
7.3.6. Proposition. Let f : M M be a homomorphism on an artinian module.
The following conditions are equivalent:
(1) f is injective
(2) f is an isomorphism
Proof. There is a number n such that Imf
n
= Imf
n+1
. For x M there is y
such that f
n
(x) = f
n+1
(y). Then f
n
(x f(y)) = 0 so x = f(y) since f is
injective. It follows that f is surjective.
7.3.7. Proposition. A nite direct sum M
1
M
n
of artinian modules M
i
is
artinian.
Proof. Use the exact sequences
0 M
1
M
n1
M
1
M
n
M
n
0
together with induction and 7.3.4.
7.3.8. Proposition. Given submodules N, L M, then the following statements
are equivalent:
(1) M/N, M/L are artinian
(2) M/N L is artinian
Proof. Use the exact sequences 3.2.7
0

M/N L

M/N M/L

M/N +L

0
together with 7.3.4.
7.3.9. Proposition. Let M be a nite and N an artinian R-module.
(1) M
R
N is artinian.
(2) Hom
R
(M, N) is artinian.
Proof. Choose R
n
M 0 exact. (1) N
n
R
n

R
N M
R
N 0 is
exact, so conclusion by 7.3.4. (2) 0 Hom
R
(M, N) Hom
R
(R
n
, N) N
n
is
exact, so conclusion by 7.3.4.
7.3.10. Proposition. Let I R be an ideal. Suppose an R/I-module M is ar-
tinian. Then M is an artinian R-module.
Proof. This is clear since a sequence of R-submodules is a sequence of R/I-
modules.
7.3.11. Proposition. Let U R be a multiplicative subset. Suppose an R-module
M is artinian. Then U
1
M is an artinian U
1
R-module.
Proof. Let i : M U
1
M. By 4.4.6 any U
1
R-submodule N U
1
M is
extended U
1
(i
1
(N)) N. So a chain is stationary.
7.3.12. Proposition. Let R = R
1
R
2
be a product of rings. An R-module
M
1
M
2
is artinian if and only if M
i
is an artinian R
i
-module.
Proof. This follows from the dictionary 2.2.9.
7.3.13. Exercise. (1) Show that a vector space is artinian if and only if it is nite di-
mensional.
7.4. ARTINIAN RINGS 107
7.4. Artinian rings
7.4.1. Denition. A ring R is an artinian ring if R is an artinian module over
itself.
7.4.2. Example. (1) A eld is artinian.
(2) A nite product of elds is artinian.
7.4.3. Proposition. (1) Let R be an artinian ring and I an ideal. Then the factor
ring R/I is artinian.
(2) A product ring R
1
R
2
is artinian if and only if each R
i
is artinian.
(3) Let R be an artinian ring and U a multiplicative subset. Then the fraction
ring U
1
R is artinian.
Proof. This follows from 7.3.10 and 7.3.11.
7.4.4. Proposition. Given ideals I, J R, then the following statements are
equivalent:
(1) R/I, R/J are artinian
(2) R/I J is artinian
Proof. This follows from 7.3.8.
7.4.5. Proposition. An artinian domain is a eld.
Proof. By 7.3.6 scalar multiplication with a nonzero element is an isomorphism.
7.4.6. Theorem. Let Rbe an artinian ring. Then all prime ideals are maximal and
there are only nitely many such.
Proof. By 7.4.5 primes are maximal. If P
i
are different maximal ideals, then
P
1
P
n+1
P
1
P
n
is a strictly decreasing chain. So there are only nitely
many maximal ideals.
7.4.7. Proposition. Let R be an artinian ring.
(1) The factor ring R/

0 is a nite product of elds.


(2) The nilradical

0 is nilpotent,

0
k
= 0 for some k.
Proof. (1) Let P
1
, . . . , P
n
be the prime and maximal ideals 7.4.6. The nilradical

0 = P
1
P
n
by 5.1.7. Conclude by Chinese remainders 1.4.3. (2)

0
k
=

0
k+1
for some k. If

0
k
,= 0 let (a) be minimal among ideals I such that
I

0
k
,= 0. By minimality (a) = (a)

0
k
, so a = ab for some b

0
k
. But b is
nilpotent 1.3.8, so a = 0 gives a contradiction. It follows, that

0
k
= 0.
7.4.8. Theorem. A ring R is artinian if and only if it has nite length.
Proof. The factor module

0
i
/

0
i+1
is an artinian module over R/

0 which by
7.4.7 is a product of elds, so it has nite length. By the exact sequences
0

0
i
/

0
i+1
R/

0
i+1
R/

0
i
0
and induction follows that R/

0
k
is artinian for all k. Since

0 is nilpotent 7.4.7
it follows that R has nite length.
7.4.9. Corollary. Let R be an artinian ring and M a module. The following con-
ditions are equivalent:
108 7. MODULES OF FINITE LENGTH
(1) M has nite length.
(2) M is nite.
(3) M is nite presented.
7.4.10. Corollary. Let R be artinian and R S a nite ring homomorphism.
(1) S is artinian.
(2) A nite length S-module N is by restriction of scalars a nite length R-
module.
(3) A nite length R-module M gives by change of rings M
R
S as nite length
S-module.
7.4.11. Proposition. Let M be an R-module of nite length.
(1) The ring R/ Ann(M) is artinian.
(2) There are only nitely many prime ideals Ann(M) P.
(3) Any prime ideal Ann(M) P is maximal.
Proof. (1) Let x M, then R/ Ann(x) Rx M is artinian. If x
1
, . . . , x
n
generate M, then Ann(M) = Ann(x
1
) Ann(x
n
). By 7.4.4 R/ Ann(M)
is artinian. (2)(3) This follows by (1) and 7.4.6.
7.4.12. Proposition. Let M be a R-module. The following conditions are equiva-
lent:
(1) M has nite length.
(2) M is nite and artinian.
Proof. (1) (2): See 7.2.10 and 7.3.3. (2) (1): Let x M, then R/ Ann(x)
Rx M is artinian. If x
1
, . . . , x
n
generate M, then Ann(M) = Ann(x
1
)
Ann(X
n
). By 7.4.8 R/ Ann(M) is artinian and a nite module over an artinian
ring has nite length.
7.4.13. Corollary. Let (R, P) be a local ring and M an artinian module. The
following conditions are equivalent:
(1) M = 0.
(2) Hom
R
(k(P), M) = 0.
Proof. A nite submodule has nite length.
7.4.14. Example. Let K be a eld. The ring R =

N
K = a : N K is not
artinian.
(1) There are maximal ideals in R
P
i
= a : N K[a(i) = 0
(2) Corresponding simple types are
R/P
i
J
i
= a : N K[a(j) = 0, i ,= j
(3) J
i
are simple ideals and the sum is an ideal

i
J
i
= J ,= R
(4) J is an ideal which has no complement in R.
7.4.15. Exercise. (1) Show that a vector space is artinian if and only if it is nite di-
mensional.
(2) Show that Z is not artinian.
7.5. LOCALIZATION 109
(3) Show that R[X] is not artinian for a nonzero ring R.
(4) Show that Q[X]/(X
2
X) is artinian.
(5) Show that a ring with nitely many ideals is artinian.
(6) Let R be a principal ideal domain and a ,= 0. Show that R/(a) is artinian.
7.5. Localization
7.5.1. Theorem. Any artinian ring is the product of nitely many artinian local
rings. Let R be artinian with maximal ideals P
1
, . . . , P
k
. Then there is n such that
P
n
1
. . . P
n
k
= 0 and
R R/P
n
1
R/P
n
k
Each R/P
n
i
is a local artinian ring.
Proof. This follows from 7.4.7, 7.4.8 and Chinese remainders 1.4.2.
7.5.2. Example. A reduced artinian ring is a nite product of elds.
7.5.3. Corollary. Let R be an artinian ring and P R a prime ideal. Then the
localization R
P
is an artinian ring.
7.5.4. Lemma. Let P R be a maximal ideal and M a module. Assume u / P
and n N.
(1) (u) +P
n
= R.
(2) Scalar multiplication u
M/P
n
M
: M/P
n
M M/P
n
M is an isomorphism.
(3) The canonical map M/P
n
M (M/P
n
M)
P
is an isomorphism.
(4) If M is nite then M/P
n
M has nite length.
7.5.5. Theorem. Let M be an R-module of nite length and Ann(M) P
1
, . . . , P
k
the maximal ideals. Then the sequence
0 P
n
i
M M M
P
i
0
is exact and M/P
n
i
M M
P
i
is a nite length R
P
i
-module. There are isomor-
phisms
M

i
M
P
i

i
M/P
n
i
M
and the length is

R
(M) =

R
P
i
(M
P
i
)
Proof. The ring R/ Ann(M) R/P
n
1
+ Ann(M) R/P
n
k
+ Ann(M) by
7.4.11 and 7.5.1. Therefore M M/P
n
1
M M/P
n
k
M and by localization
M
P
i
M/P
n
i
M.
7.5.6. Corollary. Let Rbe artinian with maximal ideals P
1
, . . . , P
k
and M a nite
module. Then the sequence
0 P
n
i
M M M
P
i
0
is exact and M/P
n
i
M M
P
i
is a nite R
P
i
-module. There are isomorphisms
M

i
M
P
i

i
M/P
n
i
M
and the length is

R
(M) =

R
P
i
(M
P
i
)
110 7. MODULES OF FINITE LENGTH
7.5.7. Proposition. Let (R, P) (S, Q) be a local ring homomorphism and as-
sume that k(P) k(Q) is a nite extension. If N is a nite length S-module, then
N is a nite length R-module and

R
(N) = rank
k(P)
(k(Q))
S
(N)
Proof. Reduce to N = k(Q).
7.5.8. Proposition. Let (R, P) (S, Q) be a local ring homomorphism and as-
sume that S/PS is a nite length S-module. Let M be a nite length R-module.
(1) M
R
S is a nite length S-module.
(2) In general

S
(M
R
S)
S
(S/PS)
R
(M)
(3) If R S is at then

S
(M
R
S) =
S
(S/PS)
R
(M)
Proof. The case M = k(P) is clear. Conclude by induction.
7.5.9. Exercise. (1) Let K L be a nite eld extension and W a nite vector space
over L. Show that
rank
K
(W) = rank
K
(L) rank
L
(W)
(2) Let K L be a nite eld extension and V a nite vector space over K. Show that
rank
L
(V
K
L) = rank
K
(V )
7.6. Local artinian ring
7.6.1. Lemma. Let (R, P) be a local artinian ring and M any module. The fol-
lowing conditions are equivalent:
(1) M = 0.
(2) Hom
R
(k(P), M) = 0.
Proof. (2) (1): A nonzero submodule Rx M has nite length and contains a
simple submodule k(P) L M. So Hom
R
(k(P), M) ,= 0.
7.6.2. Proposition. Let (R, P) be a local artinian ring and k(P) E an injective
envelope.
(1) There are isomorphisms
k(P) Hom
R
(k(P), k(P)) Hom
R
(k(P), E)
(2) For a nite module M, the module Hom
R
(M, E) has nite length and

R
(Hom
R
(M, E)) =
R
(M)
Proof. (1) A nonzero homomorphism f : k(P) E has Imf = k(P) since
the extension is essential. (2) An exact sequence 0 N M k(P) 0
7.1.4 gives an exact sequence with Hom
R
(, E). Conclude by (1) and induction
on
R
(M).
7.6.3. Corollary. Let (R, P) be a local artinian ring and k(P) E an injective
envelope. Then E has nite length

R
(E) =
R
(R)
7.6. LOCAL ARTINIAN RING 111
7.6.4. Theorem. Let (R, P) be a local artinian ring and k(P) E an injective
envelope. There is a natural isomorphism for any nite module M
x ev
x
: M Hom
R
(Hom
R
(M, E), E)
Proof. The case M = k(P) is clear by 7.6.2. Let 0 N M L 0 be a
short exact sequence. Then the following diagram has exact rows.
0

N

0
0

Hom
R
(Hom
R
(N, E), E)

Hom
R
(Hom
R
(M, E), E)

Hom
R
(Hom
R
(L, E), E)

0
Conclusion by the ve lemma 3.2.8 and induction on the length.
7.6.5. Corollary. Let (R, P) be a local artinian ring and k(P) E an injective
envelope. There is an isomorphism
1
1
E
: R Hom
R
(E, E)
7.6.6. Lemma. Let (R, P) be a local artinian ring and k(P) E an injective
envelope. If E

is injective and Hom


R
(k(P), E

) is nite, then E

E
n
, where
n =
R
(Hom
R
(k(P), E

).
Proof. If n = 0 then E

= 0 by 7.6.1. For n > 0 a nonzero k(P) E

gives
an injective extension 0 E E

. Since E is injective there is a splitting


E

E E

. Conclusion by induction on n.
7.6.7. Theorem. Let (R, P) be a local artinian ring and k(P) E an injective
envelope. Let M E

be an injective envelope. Then


(1) Hom
R
(k(P), M) Hom
R
(k(P), E

)
(2) If M is artinian, then E

E
n
, where n =
R
(Hom
R
(k(P), M).
(3) An artinian M module has nite length and

R
(M)
R
(Hom
R
(k(P), M))
R
(R)
Proof. (1) A homomorphism f : k(P) E

has Imf M since the extension


is essential. (2) E

E
n
by 7.6.6 and n is determined by (1) and 7.6.2. (3) This
follows from (2) and 7.6.3.
7.6.8. Corollary. Let R be an artinian ring and M an R-module. The following
conditions are equivalent:
(1) M has nite length.
(2) M is nite.
(3) M is artinian.
Proof. Reduce to the case where R is local by 7.5.1.
7.6.9. Theorem. Let (R, P) be a local artinian ring and k(P) E an injective
envelope. The following conditions are equivalent:
(1) R is injective.
(2) R E.
(3) Hom
R
(k(P), R) k(P).
Proof. (1) (2): By 7.6.6 R E
n
and by 7.6.3 n = 1. (2) (3): Immediate
from 7.6.6. (3) (1): Let R E

be an injective envelope. By 7.6.7 R = E

.
112 7. MODULES OF FINITE LENGTH
7.6.10. Denition. A ring satisfying the conditions 7.6.9 is a local artinian Goren-
stein ring.
7.6.11. Example. Let R be a principal ideal domain and p R an irreducible
element. Then Hom
R
(R/(p), R/(p
n
)) (p
n1
)/(p
n
) R/P and R/(p
n
) is a
local artinian Gorenstein ring, 7.6.9.
7.6.12. Exercise. (1) Let p be a prime number. Show that Z/(p
k
) is a local artinian
Gorenstein ring.
8
Noetherian rings
8.1. Noetherian modules
8.1.1. Lemma. Let M be a module. The following conditions are equivalent:
(1) Any increasing chain M
i
M
i+1
. . . of submodules is stationary: there
is n such that M
i
= M
i+1
for i > n.
(2) Any nonempty subset of submodules of M contains a maximal element.
(3) Any submodule of M is nite.
Proof. (1) (2): Suppose a nonempty subset of submodules do not contain a
maximal element. Then choose a non stationary ascending chain. (2) (1): An
ascending chain containing a maximal element is stationary. (2) (3): Let N be a
submodule of M and choose a maximal element N

in the set of nite submodules


of N. For y N, the module N

+Ry is nite, so N

= N

+Ry gives N = N

nite. (3) (1): The union


j
M
j
is a submodule, generated by x
1
, . . . , x
m
M
n
,
so M
i
=
j
M
j
, i > n.
8.1.2. Denition. A module M which satises the conditions of 8.1.1 is a noe-
therian module.
8.1.3. Theorem. Let
0

M
f

N
g

L

0
be an exact sequence of modules over the ring R. The following statements are
equivalent:
(1) N is noetherian.
(2) M and L are noetherian.
Proof. (1) (2): A chain M
i
in M gives a chain f(M
i
) in N which becomes
stationary. So M is noetherian. A chain in L
i
in L gives a chain in g
1
(L
i
)
in N, which becomes stationary. Then also the original chain L
i
= g(g
1
(L
i
))
becomes stationary and L is noetherian. (2) (1): A chain N
i
in N, induces
chains f
1
(N
i
) in N and g(N
i
) in L which become stationary. By the snake
lemma 3.2.4
f
1
(N
i
)/f
1
(N
i1
) N
i
/N
i1
g(N
i
)/g(N
i1
)
is exact and the original chain becomes stationary.
8.1.4. Corollary. (1) Let M N be a submodule. Then N is noetherian if and
only if both M and N/M are noetherian.
(2) Let f : M N be a homomorphism.
(a) M is noetherian if and only if Ker f, Imf are noetherian.
(b) N is noetherian if and only if Imf, Cok f are noetherian
113
114 8. NOETHERIAN RINGS
Proof. Use the sequences 3.1.8.
8.1.5. Proposition. Let f : M M be a homomorphism on a noetherian module.
The following conditions are equivalent:
(1) f is surjective
(2) f is an isomorphism
Proof. Analog of the proof of 7.3.4. There is a number n such that Ker f
n+1
=
Ker f
n
. For x Ker f there is y such that x = f
n
(y), since f is surjective.
Then f
n+1
(y) = f(x) = 0, so y Ker f
n+1
= Ker f
n
. Then x = f
n
(y) = 0
and f is injective.
8.1.6. Proposition. A nite direct sum M
1
M
n
of noetherian modules M
i
is noetherian.
Proof. Use the exact sequences
0 M
1
M
n1
M
1
M
n
M
n
0
together with induction and 8.1.3.
8.1.7. Proposition. Given submodules N, L M, then the following statements
are equivalent:
(1) M/N, M/L are noetherian
(2) M/N L is noetherian
Proof. Use the exact sequence 3.2.7
0

M/N L

M/N M/L

M/N +L

0
together with 8.1.3.
8.1.8. Proposition. Let M be a nite and N a noetherian R-module.
(1) M
R
N is noetherian.
(2) Hom
R
(M, N) is noetherian.
Proof. Choose R
n
M 0 exact. (1) N
n
R
n

R
N M
R
N 0 is
exact, so conclusion by 8.1.3. (2) 0 Hom
R
(M, N) Hom
R
(R
n
, N) N
n
is
exact, so conclusion by 8.1.3.
8.1.9. Proposition. Let I R be an ideal. Suppose an R/I-module M is noe-
therian. Then M is a noetherian R-module.
Proof. This is clear since a chain of R-submodules is a chain of R/I-modules.
8.1.10. Proposition. Let U R be a multiplicative subset. Suppose an R-module
M is noetherian. Then U
1
M is noetherian U
1
R-module.
Proof. Let i : M U
1
M. By 4.4.6 any U
1
R-submodule N U
1
M is
extended U
1
(i
1
(N)) N. So a chain is stationary.
8.1.11. Proposition. Let R = R
1
R
2
be a product of rings. An R-module
M
1
M
2
is noetherian if and only if M
i
is an noetherian R
i
-module.
Proof. This follows from 8.1.9 and 8.1.3.
8.1.12. Proposition. Let M be a R-module. The following conditions are equiva-
lent:
8.2. NOETHERIAN RINGS 115
(1) M has nite length.
(2) M is noetherian and artinian.
Proof. (1) (2): A chain produces a monoton bounded sequence of lengths. (2)
(1): M is nite and artinian, so conclusion by 7.4.12.
8.1.13. Example. (1) A vector space is noetherian if and only if it is nite.
8.1.14. Exercise. (1) Show that

N
Z is not noetherian.
(2) Show that

N
Z is not noetherian.
(3) Show that Qis not a noetherian Z-module.
(4) Let p be a prime number. Show that the Z
(p)
-submodules of Q containing Z
(p)
is
of form p
n
Z
(p)
and conclude that Q/Z
(p)
is an artinian but not noetherian Z
(p)
-
module.
8.2. Noetherian rings
8.2.1. Denition. A ring R is a noetherian ring if R is a noetherian module.
8.2.2. Proposition. The following conditions are equivalent:
(1) R is noetherian.
(2) Any ideal is nite.
(3) Any increasing sequence of ideals is stationary.
(4) Any nonempty subset of ideals of contains an ideal maximal for inclusion.
(5) Any ideal is noetherian.
Proof. This follows from 8.1.1.
8.2.3. Proposition. Let R be a noetherian ring.
(1) The nilradical is nilpotent. For some n

0
n
= 0
(2) Some power of the radical of an ideal is contained in the ideal. For some n

I
n
I
Proof. (1)

0 = (b
1
, . . . , b
m
) such that b
k
i
= 0. Then (

a
i
b
i
)
mk
= 0. (2) Use
(1) on R/I.
8.2.4. Proposition. (1) Let I R be an ideal in a noetherian ring. Then the
factor ring R/I is a noetherian ring.
(2) The product ring R
1
R
2
is noetherian if and only if each R
i
is noetherian.
(3) Let U R be a multiplicative subset in a noetherian ring. Then the fraction
ring U
1
R is a noetherian ring.
Proof. Use 8.1.3 and 8.1.10.
8.2.5. Proposition. Given ideals I, J R, then the following statements are
equivalent:
(1) R/I, R/J are noetherian
(2) R/I J is noetherian
Proof. This is a special case of 8.1.7.
8.2.6. Theorem. Let R be a noetherian ring. Then any nite R-module is noether-
ian.
116 8. NOETHERIAN RINGS
Proof. Let M be a nite R-module. There is a surjective homomorphism R
n

M 0, 6.1.2. Conclusion by 8.1.3.


8.2.7. Corollary. Let R be a noetherian ring. Any nite R-module is nite pre-
sented.
8.2.8. Proposition. Let M be an R-module. The following Statements are equiva-
lent:
(1) M is a noetherian.
(2) R/ Ann(M) is a noetherian ring and M is nite.
Proof. (1) (2): Let x M, then R/ Ann(x) Rx M is noetherian. If
x
1
, . . . , x
n
generate M, then Ann(M) = Ann(x
1
) Ann(X
n
). By 8.2.4
R/ Ann(M) is noetherian. (2) (1): By 8.2.6 M is a nite R/ Ann(M)-module.
Conclusion by 8.1.9.
8.2.9. Proposition. Let R be a noetherian ring and M, N noetherian modules.
(1) M
R
N is noetherian.
(2) Hom
R
(M, N) is noetherian.
Proof. Conclusion by 6.1.8, 6.1.9, 8.2.6.
8.2.10. Theorem. Let R be a noetherian ring and U a multiplicative subset. For
a nite module M and any module N the homomorphism
U
1
Hom
R
(M, N) Hom
U
1
R
(U
1
M, U
1
N)
is an isomorphism.
Proof. Conclusion by 6.5.10, 8.2.7.
8.2.11. Proposition. Let R be a noetherian ring. If E is an injective R-module,
then U
1
E is an injective U
1
R-module.
Proof. Let I R be an ideal. Then Hom
R
(R, E) Hom
R
(I, E) 0 is exact.
By 8.2.10 Hom
U
1
R
(U
1
R, U
1
E) Hom
U
1
R
(U
1
I, U
1
E) 0 is exact.
So U
1
E is injective 3.6.8, using that any ideal is extended 4.3.6.
8.2.12. Theorem. For a ring R, the following statements are equivalent:
(1) R is a noetherian ring.
(2) For any family E

of injective modules, the sum

is injective.
Proof. (1) (2): Let I R be an ideal. A homomorphism f : I

E

has Imf contained in a nite sum, which is injective 3.6.7. So f extends to R

, giving injectivity. (2) (1): Let I


n
R be an increasing chain of ideals.
Put I = I
n
and choose an injective envelope I/I
n
E
n
. The homomorphism
f : I

E
n
extends to f

: R

E
n
. Since Imf is contained in a nite
sum, I/I
n
= 0 for n >> 0.
8.2.13. Theorem. Let R be a noetherian ring and F

a family of at modules.
Then the product

is at.
Proof. An ideal I R is nite presented. By 6.5.8 the homomorphsim I
R

is a product of injective homomorphisms

(I
R
F

,
which is injective giving the conclusion.
8.3. FINITE TYPE RINGS 117
8.2.14. Theorem. If R is a ring such that every prime ideal is nite, then it is
noetherian.
Proof. If R is not noetherian, then the set of not nite ideals is nonempty and by
Zorns lemma it has a maximal element I. Since I is not prime there is a, b / I
and ab I. The ideals I + (a) and I : (a) are both greater that I and therefore
nite. Let I + (a) = (a
1
, . . . , a
m
, a), a
i
I and I : (a) = (b
1
, . . . , b
n
). Assume
c = c
1
a
1
+ +c
m
a
m
+da I. Then d I : (a) and d = d
1
b
1
+ +d
n
b
n
. So
I = (a
1
, . . . , a
m
, ab
1
, . . . , ab
n
) is nite. It follows that R must be noetherian.
8.2.15. Exercise. (1) Show that a principal ideal domain is noetherian.
(2) Let I R be an ideal in a noetherian ring. Show that R/I is noetherian.
(3) Let K be a eld and R = K[X
1
, X
2
, . . . ] the polynomial ring in countable many
variables. Show that R is not noetherian.
(4) Let K be a eld. Show that the ring R =

N
K is not noetherian.
8.3. Finite type rings
8.3.1. Theorem (Hilberts basis theorem). Let R be a noetherian ring. Then the
ring of polynomials R[X] is noetherian.
Proof. Assume I R[X] to be a not nite ideal. Choose a sequence f
1
, f
2
,
I such that
f
i
= a
i
X
d
i
+ terms of lower degree , a
i
,= 0
and f
i+1
has lowest degree in I(f
1
, . . . , f
i
). The ideal of leading coefcients is
nitely generated by a
1
, . . . , a
n
. Then a
n+1
= b
1
a
1
+ +b
n
a
n
and d
1

d
n+1
= d gives
f
n+1
b
1
X
dd
1
f
1
b
n
X
dd
n
f
n
in I(f
1
. . . , f
n
) of degree less than d. By contradiction the ideal I is nite.
8.3.2. Corollary. Let R be a noetherian ring.
(1) A polynomial ring R[X
1
, . . . , X
n
] in nitely many variables is noetherian.
(2) If R S a nite type ring over R, then S is noetherian.
8.3.3. Corollary. Let R be a noetherian ring and M a nite module. Then the ring
R M, 2.1.14, is noetherian.
Proof. Let x
1
, . . . , x
n
generate M. Then R[X
1
, . . . , X
n
] R M, X
i
x
i
is
surjective. So R R M is of nite type.
8.3.4. Example. Let I R be an ideal in a noetherian ring. Then there are noe-
therian rings.
(1) G
I
R =
n0
I
n
/I
n+1
.
(2) B
I
R =
n0
I
n
.
8.3.5. Theorem (Krulls intersection theorem). Let I be an ideal in a noetherian
ring R. Then there is a I such that
1 +a Ann(

n
I
n
)
118 8. NOETHERIAN RINGS
Proof. Let I = (u
1
, . . . , u
m
). If b I
n
then b = f
n
(u
1
, . . . , u
m
) where f
n

R[X
1
, . . . , X
m
] are homogeneous of degree n. By Hilberts basis theorem, 8.3.1
there is N such that
f
N+1
= f
1
g
1
+ +f
N
g
N
where g
n
is homogeneous of degree N n + 1 > 0. By substitution b bI and
I(I
n
) = I
n
. Now I
n
R is nite, so conclusion by 6.3.4.
8.3.6. Corollary. Let I be an ideal in a noetherian ring R such that the elements
1 +a, a I are nonzero divisors. Then

n
I
n
= 0
8.3.7. Corollary. Let I be an ideal in a noetherian ring R and M a nite module.
Then there is a I such that
1 +a Ann(

n
I
n
M)
Proof. Use 8.3.5 on the ring R M and the ideal I M. Clearly (I + M)
n
=
I
n
+I
n1
M.
8.3.8. Corollary. Let I be an ideal in a noetherian ring R and M a nite module
such that the elements 1 +a, a I are nonzero divisors on M. Then

n
I
n
M = 0
8.3.9. Theorem. If R S be a nite extension. Then R is noetherian if and only
if S is noetherian.
Proof. Assume S is noetherian. Let E

be a family of injective R-modules. Then

Hom
R
(S, E

) is an injective S-module. Since S is nite over R, there is an


isomorphism 6.1.14

Hom
R
(S, E

) Hom
R
(S,

). By 6.6.8

E

is
injective and by 8.2.12 R is noetherian.
8.3.10. Example. Let K be a eld and R = K[X
1
, X
2
, . . . ]/(X
1
x
2
X
3
, x
2

x
3
X
4
, . . . ). Put P = (X
1
, X
2
, . . . ).
(1) P is maximal
(2) P
2
= P.
(3) P(
n
P
n
) =
n
P
n
.
8.3.11. Exercise. (1) Show that if R[X] is noetherian, then R is noetherian.
(2) Show that the subring Z[2X, 2X
2
, . . . ] Z[X] is not noetherian. Conclude that the
extension is not nite.
8.4. Power series rings
8.4.1. Theorem. Let R be a noetherian ring. Then the power series ring R[[X]] is
noetherian.
Proof. Let P R[[X]] be a prime ideal. Then P +(X)/(X) = (a
1
, . . . , a
n
) R
is a nite ideal. If X P then P = (a
1
, . . . , a
n
, X) is nite. Suppose X / P
and choose f
i
= a
i
+ terms of positive degree P. If g P then write g =
b
10
f
1
+ + b
n0
f
n
+ Xg
1
for some b
i0
R. Since P is prime, g
1
P. Now
g
1
= b
11
f
1
+ +b
n1
f
n
+Xg
2
and so on. Put h
i
=

k
b
ik
X
k
, then g =

i
h
i
f
i
.
So P = (h
1
, . . . , h
n
) is nite and R[[X]] is noetherian by 8.2.14.
8.4. POWER SERIES RINGS 119
8.4.2. Corollary. Let R be a noetherian ring.
(1) A power series ring R[[X
1
, . . . , X
n
]] in nitely many variables is noetherian.
(2) Let I R[[X
1
, . . . , X
n
]] be an ideal, then the factor ring R[[X
1
, . . . , X
n
]]/I
is noetherian.
8.4.3. Proposition. Let R be a principal ideal domain. Then the power series ring
R[[X]] is a unique factorization domain.
Proof. By 5.1.12 it is enough to show that a nonzero prime contains a principal
prime. Let P R[[X]] be nonzero prime ideal. If X P then (X) P is
a principal prime. Suppose P ,= (X) and P + (X)/(X) = (a) R. Choose
f = a + terms of positive degree P. If g P then g = b
0
f + Xg
1
for
some b
0
R. Since P is prime, g
1
P. Now g
1
= b
1
f + Xg
2
and so on. Put
h =

k
b
k
X
k
, then g = hf. So P = (f) is a principal prime itself.
8.4.4. Proposition. Let I R be an ideal in a noetherian ring. Then there is a
canonical isomorphism
R[[X]]/IR[[X]] R/I[[X]]
Proof. The projection R[[X]] R[[X]]/IR[[X]] factors over R/I[[X]] since I is
nitely generated. Then there is an inverse to the homomorphism 1.9.8.
8.4.5. Corollary. If P R is a prime ideal in a noetherian ring, then PR[[X]]
R[[X]] is a prime ideal.
8.4.6. Proposition. Let R be a noetherian ring.
(1) The inclusion R[X] R[[X]] is a at homomorphism.
(2) The inclusion R R[[X]] is a faithfully at homomorphism.
Proof. (1) Let I R[X] be an ideal and let

a
i
f
i
K = Ker(I
R[X]
R[[X]] R[[X]]. Then

a
i
f
i
= 0. Chase the diagram
I
R[X]
(X
n
)

I
R[X]
R[[X]]

R[[X]]

I
R[X]
R[[X]]/(X
n
)

R[[X]]/(X
n
)
Since R[X]/(X
n
) = R[[X]]/(X
n
) it follows that

a
i
f
i
=

a
i
f
i
1 = 0
I
R[X]
R[[X]]/(X
n
). Therefore

a
i
f
i
X
n
(I
R[X]
R[[X]]). It follows that
K

n
X
n
(I
R[X]
R[[X]]). I
R[X]
R[[X]] is a nite R[[X]]-module and 1+aX
is a unit, so conclusion by 8.3.8. (2) The homomorphism R R[X] R[[X]]
is a composition of at homomorphisms by (1). For any maximal ideal P R
the homomorphism R
P
R
P
[[X]] is local. It follows that P = Q R for some
Q R[[X]], so faithfully at by 5.6.15.
8.4.7. Exercise. (1) Show that if R[[X]] is noetherian, then R is noetherian.
120 8. NOETHERIAN RINGS
8.5. Localization of noetherian rings
8.5.1. Theorem. Let R be a noetherian ring and P a prime ideal. Then the local
ring R
P
is a noetherian ring.
Proof. Any ideal is extended 4.3.6 or just 8.2.4.
8.5.2. Theorem. Let R be a noetherian ring and P a prime ideal. For a nite
module M and any module N the homomorphism
Hom
R
(M, N)
P
Hom
R
P
(M
P
, N
P
)
is an isomorphism.
Proof. A special case of 8.2.10.
8.5.3. Proposition. Let R be a noetherian ring and M a nite module. Let P be a
prime ideal. Then R
P
is a noetherian ring and M
P
is a nite R
P
-module.
8.5.4. Theorem (Krulls intersection theorem). Let (R, P) be a noetherian local
ring and M a nite module. Then

n
P
n
M = 0
Proof. From 8.3.7, (1 + a)

n
P
n
M = 0 for some a P. Now use that 1 + a is
a unit.
8.5.5. Corollary. Let (R, P) be a noetherian local ring and I P an ideal. Then

n
I
n
= 0
8.5.6. Theorem. Let (R, P) be a noetherian local ring and F a nite module. The
following conditions are equivalent:
(1) F is free.
(2) F is projective.
(3) F is at.
(4) P
R
F F is injective.
Proof. This follows from 6.5.14 as nite modules are nite presented.
8.5.7. Proposition. Let R be a noetherian ring and F a nite module. The follow-
ing conditions are equivalent:
(1) F is projective.
(2) F is at.
(3) P
R
F F is injective for all maximal ideals P.
(4) F
P
is free for all maximal ideals P.
Proof. This follows from 6.5.16 as nite modules are nite presented.
8.5.8. Exercise. (1) Is it true that if U
1
R is noetherian, then R is noetherian?
8.6. PRIME FILTRATIONS OF MODULES 121
8.6. Prime ltrations of modules
8.6.1. Proposition. Let R be a ring and M ,= 0 a nonzero module. An ideal
Ann(x) maximal in the set of ideals Ann(y)[0 ,= y M is a prime ideal.
Proof. Let Ann(x) be a maximal annihilator. Suppose a, b R such that ab
Ann(x) and b / Ann(x). Then
Ann(x) Ann(bx) ,= R
Consequently Ann(x) = Ann(bx), in particular a Ann(x).
8.6.2. Corollary. Let R be a noetherian ring and M ,= 0 a nonzero module. Then
there is x M such that Ann(x) is a prime ideal.
8.6.3. Theorem. Let R be a noetherian ring and M ,= 0 a nite R-module. Then
there exists a nite ltration of M by submodules
0 = M
0
M
1
M
n1
M
n
= M
such that M
i
/M
i1
, i = 1, . . . , n is isomorphic to an R-module of the form R/P
i
where P
i
is a prime ideal in R.
Proof. The set of submodules of M for which the theorem is true is nonempty by
8.2.6. Let N M be maximal in this set. Suppose N ,= M. By 8.2.6 applied to
M/N there is a chain N N

M such that N

/N is isomorphic to an R-module
of the form R/P

where P

is a prime ideal. This contradicts the maximality of N.


So N = M.
8.6.4. Corollary. Let R be a nonzero noetherian ring. Then there exists a nite
ltration of ideals
0 = I
0
I
1
I
n1
I
n
= R
such that I
i
/I
i1
, i = 1, . . . , n is isomorphic to an R-module of the form R/P
i
where P
i
is a prime ideal in R.
8.6.5. Example. In Z there is a ltration
0 (p
n
) (p
n1
) (p) Z
of any length with factors (p
n
) Z and (p
i1
)/(p
i
) Z/(p) for any prime
number p.
8.6.6. Theorem. Let R be a ring. The following statements are equivalent:
(1) R is artinian.
(2) R is noetherian and all prime ideals are maximal.
Proof. (1) (2): By 7.4.8 an artinian ring has nite length and therefore noether-
ian. Primes are maximal by 7.3.11. (2) (1): By 8.6.4 there is a nite composition
series.
8.6.7. Corollary. Let R be a ring and M a module. The following statements are
equivalent:
(1) M has nite length.
(2) R/ Ann(M) is artinian and M is nite.
8.6.8. Theorem. If R S be a nite extension. Then R is artinian if and only if
S is artinian.
122 8. NOETHERIAN RINGS
Proof. Suppose S is artinian. By 8.3.9 R is noetherian. By 6.6.5 any prime ideal
in R is maximal. Conclusion by 8.6.6.
8.6.9. Proposition. Let R be a noetherian ring. The number of minimal prime
ideals is nite.
Proof. Choose a ltration 8.6.4, 0 = I
0
I
n
= R with I
i
/I
i1
R/P
i
,
where P
i
is a prime ideal. Let P be a minimal prime ideal in R. Then (I
i
/I
i1
)
P

(R/P
i
)
P
,= 0 if and only if P
i
P. Thus P = P
i
for some i since R
P
,= 0.
8.6.10. Proposition. Let R be a noetherian ring such that the local rings R
P
are
domains for all maximal ideals P. Then R is a nite product of domains.
Proof. Let P
1
, . . . , P
n
be the minimal primes 8.6.9. The intersection is 0 since the
ring is reduced 5.4.10. They are comaximal since a domain has a unique minimal
prime. Conclusion by Chinese remainders 1.4.2.
8.6.11. Exercise. (1) Compute a ltration 8.6.3 of the Z-module Z/(36).
9
Primary decomposition
9.1. Zariski topology
9.1.1. Denition. Let R be a ring.
(1) The set of prime ideals is the spectrum and denoted X = Spec(R).
(2) For a ring homomorphism : R S restriction denes the associated map
a
: Spec(S) Spec(R)
Q
1
(Q)
(3) For a subset B R
V (B) = P Spec(R)[B P
is a subset of the spectrum.
(4) For a subset B R
X
B
= Spec(R)V (B)
is the complement.
9.1.2. Lemma. Let R be a ring.
(1) For a subset B R
V (B) = V (RB) = V (

RB)
So V (B) depends only on the radical of the ideal generated by B.
(2) For a subset B R
V (B) =

bB
V (b)
(3) For a subset B R
X
B
=
_
bB
X
b
9.1.3. Lemma. Let R be a ring.
(1) Spec(R) = V (0), = V (1).
(2) V (B
1
) V (B
2
) = V (B
1
B
2
).
(3)

V (B

) = V (

).
Proof. (2) Clearly V (B
i
) V (B
1
B
2
). If P / V (B
i
) then choose b
i
B
i
P.
The product b
1
b
2
/ P, so P / V (B
1
B
2
).
9.1.4. Denition. Let R be a ring.
(1) The subsets V (B), B R are the closed sets in the Zariski topology on
Spec(R).
(2) The subsets X
B
, B Rare the open sets in the Zariski topology on Spec(R).
(3) Subsets X
b
, b R are the principal open subsets.
123
124 9. PRIMARY DECOMPOSITION
9.1.5. Proposition. Let R be a ring.
(1) Spec(R) = X
1
, = X
0
.
(2) X
b
1
X
b
2
= X
b
1
b
2
.
(3) Spec(R)V (B) =

bB
X
b
(4) The subset X
b
, b R is a basis for the open sets in the Zariski topology.
9.1.6. Lemma. Let R be a ring and I, J ideals. The following conditions are
equivalent:
(1) V (I) V (J).
(2) J

I.
(3)

J

I.
Proof. This is clear from

I =

IP
P, 5.1.8.
9.1.7. Proposition. Let R be a ring and X = Spec(R).
(1) V (I) = V (J) if and only if

J =

I.
(2) V (I) = X if and only if I

0.
(3) V (I) = if and only if I = R.
(4) X
b
= X if and only if b is a unit.
(5) X
b
= if and only if b

0.
Proof. This follows from 9.1.6.
9.1.8. Proposition. Let : R S be a ring homomorphism.
(1) The map
a
: Spec(S) Spec(R) is continuous.
(2) If I R is an ideal, then
a

1
(V (I)) = V (IS).
(3) If b R and X = Spec(R) then
a

1
(X
b
) = X
(b)
.
(4) If J S is an ideal, then the closure
a
(V (J)) = V (J R).
(5) If X = Spec(R), then the closure
a
(X) = V (Ker ).
Proof. (1) This follows from (2). (2) Calculate
a

1
(V (I)) = Q[
a
(Q)
V (I) = Q[I Q R = V (IS).
9.1.9. Proposition. Let R be a ring and X = Spec(R).
(1) Let I R be an ideal. Then
Spec(R/I) V (I) Spec(R)
is a homeomorphism onto the closed subset V (I).
(2) Let b R. Then
Spec(b
n

1
R) X
b
Spec(R)
is a homeomorphism onto the principal open subset X
b
.
Proof. This is a restatement of 1.3.5 and 5.1.5.
9.1.10. Theorem. Let R be a ring. Then any principal open subset X
b
is quasi-
compact. That is, any open covering X
b
=

may be rened to an nite


covering X
b
= U

1
U

n
.
Proof. Assume by 9.1.9 X
b
= X. By 9.1.6 U

= X
b

. So V (b

) =
and therefore some (b

1
, . . . , b

n
) = R. Clearly X = X
b

1
X
b

n

U

1
U

n
.
9.1. ZARISKI TOPOLOGY 125
9.1.11. Theorem. Let R be a ring. Spec(R) is connected if and only if R is not a
product of two nonzero rings.
Proof. If Ris a nontrivial product, then the are proper ideals I
1
+I
2
= R, I
1
I
2
=
0. So Spec(R) = V (I
1
) V (I
2
) is not connected. Conversely if Spec(R) =
V (I
1
)V (I
2
) is not connected, then the ideals are proper and I
1
+I
2
= R, I
1
I
2

0. Choose a
i
I
i
such that a
1
+ a
2
= 1 and n so big that a
n
1
a
n
2
= 0. Now
V (a
n
1
) V (a
n
2
) = V (a
1
) V (a
2
) = Spec(R) so for some b
i
, b
1
a
n
1
+ b
2
a
n
2
= 1
and (b
1
a
n
1
)(b
2
a
n
2
) = 0. It follows that R R/(b
1
a
n
1
) R/(b
2
a
n
2
) is a product of
nonzero rings.
9.1.12. Lemma. Let X ,= be a topological space. The following conditions are
equivalent:
(1) X is not a union of two proper closed subsets.
(2) Any two nonempty open subsets intersects nonempty.
(3) Any nonempty open subset is dense in X.
Proof. (1) (2): This is clear. (2) (3): If U is open and nonempty, then
X

UU = gives

U = X. (3) (2): If two open U
1
U
2
= , then

U
1
XU
2
.
So U
1
and U
2
cannot both be nonempty.
9.1.13. Denition. An irreducible space is a topological space satisfying the con-
ditions in 9.1.10. A subset of a topological space is an irreducible subset if it is
an irrecucible space in the induced topology. A maximal irreducible subset is an
irreducible component.
9.1.14. Lemma. Let X ,= be a topological space.
(1) If X is irreducible, then any nonempty open subset is irreducible.
(2) If Y X is an irreducible subset, then the closure

Y is irreducible.
(3) If X is irreducible, then X is connected.
(4) IfX is irreducible and f : X Y is continuous. Then f(X) is an irreducible
subset.
(5) Any irreducible component in X is closed.
(6) Any irreducible subset in X is contained in an irreducible component.
(7) X is the union of irreducible components.
Proof. Do easy topology homework.
9.1.15. Theorem. Let R be a ring.
(1) Spec(R) is irreducible if and only if the nilradical

0 is a prime ideal.
(2) A closed subset V (I) is irreducible if and only if the radical

I is a prime
ideal.
(3) An irreducible component is of the form V (P) where P is a minimal prime
ideal.
Proof. Let X = Spec(R). (1) If X is irreducible and a
1
, a
2
/

0, then V (a
1
a
2
) =
V (a
1
) V (a
2
) ,= X. So a
1
a
2
/

0. Conversely, if

0 is a prime, then replace


with R/

0 and assume R is a domain. If V (I


1
I
2
) = V (I
1
) V (I
2
) = X then
I
1
I
2
= 0. It follows that I
1
= 0 or I
2
= 0. (2) Use (1) on R/I. (3) From 5.1.10
any prime ideal contains a minimal prime ideal.
9.1.16. Proposition. Let R be a ring. The following conditions are equivalent:
126 9. PRIMARY DECOMPOSITION
(1) Spec(R) is a discrete topological space.
(2) R/

0 is a nite product of elds.


Proof. (1) (2): By 9.1.10 Spec(R) is nite and consists of maximal ideals.
Conclusion by Chinese remainders 1.4.3.
9.1.17. Exercise. (1) Let K be a eld. Showthat Spec(K[X, Y ]/(XY )) is connected,
but not irreducible.
9.2. Support of modules
9.2.1. Denition. Let R be a ring and M a module.
(1) The support of M is
Supp(M) = P Spec(R)[M
P
,= 0
(2) A minimal prime ideal in Supp(M) is a minimal prime of M.
9.2.2. Proposition. Let I R be an ideal. Then as R-module
Supp(R/I) = V (I)
Proof. This is a restatement of 1.3.5.
9.2.3. Proposition. Let
0 M N L 0
be a short exact sequence of modules. Then
Supp(N) = Supp(M) Supp(L)
Proof. From 5.4.6 follows that 0 M
P
N
P
L
P
0 is exact.
9.2.4. Corollary. (1) Let N M be a submodule. Then
Supp(M) = Supp(N) Supp(M/N)
(2) Given submodules N, L M. Then
(a) Supp(M/N L) = Supp(M/N) Supp(M/L) .
(b) Supp(M/N +L) Supp(M/N) Supp(M/L) .
Proof. (2) Use the exact sequence 3.2.7
0

M/N L

M/N M/L

M/N +L

0
together with 9.2.3.
9.2.5. Theorem. Let R be a ring and M a module.
(1) M = 0 if and only if Supp(M) = .
(2) Let M be a module and P Supp(M), then V (P) Supp(M).
(3) For any module
Supp(M) V (Ann(M))
(4) If M is nite, then the support is a closed subset in the Zariski topology
Supp(M) = V (Ann(M))
Proof. (1) See 5.4.1. (2) If P Q, then M
P
= (M
Q
)
PR
Q
. (3) If M
P
,= 0 then
for u / P there is x M such that ux ,= 0. So Ann(M) P. (4) Let x
1
, . . . , x
n
generate M. If Ann(M) = Ann(x
i
) P then some Ann(x
i
) P, so
x
i
1
,= 0
in M
P
.
9.2. SUPPORT OF MODULES 127
9.2.6. Proposition. Let N
1
, . . . , N
k
M be submodules such that Supp(M/N
i
)
Supp(M/N
j
) = , i ,= j. Then the homomorphism
M/
i
N
i

i
M/N
i
is an isomorphism.
Proof. By 3.2.7
0

M/
k
1
N
i

M/
k1
1
N
i
M/N
k

M/
k1
1
N
i
+N
k

0
is exact. By 9.2.4 and induction on k
Supp(M/
k1
1
N
i
+N
k
) (
k1
1
Supp(M/N
i
)) Supp(M/N
k
)
So the support of the cokernel is empty. Conclusion by 9.2.5.
9.2.7. Proposition. Let R be a ring and M, N modules.
(1)
Supp(M
R
N) Supp(M) Supp(N)
(2) If M, N are nite, then
Supp(M
R
N) = Supp(M) Supp(N)
(3) If M is nite, then
Supp(Hom
R
(M, N)) Supp(M) Supp(N)
Proof. There is an isomorphism (M
R
N)
P
M
P

R
P
N
P
. (1) This is clear.
(2) This follows from 6.4.3. (3) By 6.5.10 there is an injective homomorphism
0 Hom
R
(M, N)
P
Hom
R
P
(M
P
, N
P
).
9.2.8. Proposition. Let (R, P) (S, Q) be a local homomorphismand M a nite
R-module. If Supp(M) ,= then Supp(M
R
S) ,= .
Proof. Calculate
M
R
S
S
k(Q) M
R
k(P)
k(P)
k(Q)
and conclude by Nakayamas lemma 6.4.1.
9.2.9. Corollary. Let : R S be a ring homomorphism and M an R-module.
(1) For the change of rings S-module
Supp(M
R
S)
a

1
(Supp(M))
(2) If M is nite, then
Supp(M
R
S) =
a

1
(Supp(M))
9.2.10. Corollary. Let Rbe a ring, I an ideal in Rand M a nite R-module. Then
a
: Spec(R/I) Spec(R) denes a bijective correspondence
Supp(M/IM) Supp(M) V (I) = V (Ann(M) +I)
Q P = Q R
128 9. PRIMARY DECOMPOSITION
9.2.11. Corollary. Let Rbe a ring, U a multiplicative subset and M an R-module.
Then : R U
1
R,
a
: Spec(U
1
R) Spec(R) denes a bijective correspon-
dence
Supp(U
1
M) Supp(M) P Spec(R)[P U =
Q Q R
Proof. Let Q U
1
R be a prime and P = Q R. Then U
1
M
Q
M
P
gives
the result.
9.2.12. Proposition. Let M be a nite R-module and P Supp(M). Then there
is a nonzero homomorphism M R/P, that is Hom
R
(M, R/P) ,= 0.
Proof. By 7.1.4 there is a surjective homomorphism f : M
P
k(P) 0. Let
x
1
, . . . , x
n
generate M and choose u RP such that uf(x
i
) R/P. The
composite of M M
P
with uf is a nonzero homomorphism M R/P.
9.2.13. Proposition. Let R be a noetherian ring and M a nite module. M has
nite length if and only if Supp(M) consists only of maximal ideals.
Proof. By 9.2.5 Supp(M) = V (Ann(M)), so conclusion by 8.6.7.
9.2.14. Denition. A ring R with only nitely many maximal ideals is a semi-
local ring.
9.2.15. Proposition. Let R be a semi-local noetherian ring and F a nite module.
If F
P
is free of rank m for all maximal ideals, then F is free of rank m.
Proof. Let P
1
, . . . , P
k
be the maximal ideals. If R R/P
n
1
R/P
n
k
is
artinian, then by 7.5.5 F F
P
1
F
P
k
(R/P
n
1
)
m
(R/P
n
k
)
m
is free
of rank m. In general choose x
1
, . . . , x
m
F giving a basis for F/P
1
P
k
F.
as R/P
1
P
k
-module. The homomorphism R
m
F, e
i
x
i
is surjective by
Nakayamas lemma 6.4.6 and the support of the kernel is empty, so the homomor-
phism is injective by 9.2.5.
9.2.16. Exercise. (1) Let R = Z, M = Q and N = Z/(p). Show that Supp(M
R
N) ,= Supp(M) Supp(N).
9.3. Ass of modules
9.3.1. Denition. Let M be an R-module. A prime ideal P R is an associated
prime ideal of M if P = Ann(x) for some x M. The set of prime ideals
associated to M is Ass(M).
9.3.2. Proposition. Let R be a ring.
(1) Let M be a module and P a prime. P Ass(M) if and only if there is an
injective homomorphism R/P M.
(2) For any module M
Ass(M) Supp(M)
(3) Let 0 ,= N R/P be a nonzero submodule, then
Ass(N) = P
(4) Let I be an ideal and M an R/I-module. The inclusion Spec(R/I)
Spec(R) identies Ass(M) over the rings R/I and R.
9.3. ASS OF MODULES 129
Proof. This is clear from the denition. (1) R/P Rx M. (2) 0 R/P
M gives 0 k(P) M
P
. (3) P = Ann(x) of any nonzero x R/P. (4)
R/P = (R/I)/(P/I) gives by (1) identied associated primes.
9.3.3. Proposition. Let
0 M N L 0
be a short exact sequence of modules. Then
Ass(M) Ass(N) Ass(M) Ass(L)
Proof. The left inclusion is trivial. Assume M N and L = N/M. Next let
P Ass(N) such that P / Ass(M). Choose a submodule K N such that
K R/P. Then Ass(K M) Ass(K) Ass(M). It follows 9.3.2 that
K M = 0. Therefore K K/K M K + M/M N/M gives P
Ass(K) Ass(L).
9.3.4. Corollary. Let 0 M N L 0 be a split exact sequence of
modules.
Ass(N) = Ass(M) Ass(L)
9.3.5. Corollary. (1) Let N M be a submodule. Then
Ass(N) Ass(M) Ass(N) Ass(M/N)
(2) Let M, N be modules. Then
Ass(M N) = Ass(M) Ass(N)
(3) Given submodules N, L M. Then
Ass(M/N L) Ass(M/N) Ass(M/L) Ass(M/N L) Ass(M/N +L)
Proof. (3) Use the exact sequence 3.2.7
0

M/N L

M/N M/L

M/N +L

0
together with 9.3.3.
9.3.6. Proposition. Let M be a module and T Ass(M). Then there is a sub-
module N M such that
Ass(N) = T , Ass(M/N) = Ass(M)T
Proof. Choose by Zorns lemma N maximal in the set of submodules N

M
for which Ass(N

) T. Let Q Ass(M/N) and choose N L M with


L/N R/Q. Then Ass(L) T Q. By maximality of N follows that
Ass(L) , T. So Q / T and Q Ass(L) Ass(M).
9.3.7. Theorem. Let R be a noetherian ring and M a module. The following
conditions are equivalent:
(1) M = 0.
(2) Supp(M) = .
(3) Ass(M) = .
(4) M
P
= 0 for all prime P Ass(M).
Proof. (4) (1): If M ,= 0, then by 8.6.2 and 9.3.2 there is P Ass(M)
Supp(M).
130 9. PRIMARY DECOMPOSITION
9.3.8. Corollary. Let R be a noetherian ring and f : M N a homomorphism.
The following conditions are equivalent:
(1) f is injective.
(2) f
P
is injective for all prime ideals P Ass(M).
Proof. (2) (1): Ker f
P
= 0 for all P Ass(Ker f) Ass(M). Conclusion by
9.3.7.
9.3.9. Theorem. Let R be a noetherian ring and M a module.
(1) a R is a nonzero divisor on M if and only if a /
PAss(M)
P.
(2) The set of zero divisors on M is

PAss(M)
P
Proof. a
M
is injective if and only if a
M
P
is injective for all P Ass(M), 9.3.8.
This happens when a /
PAss(M)
P.
9.3.10. Proposition. Let U R be a multiplicative subset and M a module. De-
note : R U
1
R and
a
: Spec(U
1
R) Spec(R).
(1) The map
a

1
denes an inclusion
Ass(M) P Spec(R)[P U = Ass(U
1
M)
P PR
P
(2) If R is noetherian, then the map (1) is a bijective correspondence.
Proof. Let P be a prime ideal. For any homomorphism R/P M there is a
commutative diagram
R/P

U
1
R/P

U
1
M
(1) If P Ass(M) and P U = then U
1
P Ass(U
1
M). (2) If U
1
P
Ass(U
1
M), then PU = . 8.2.9 gives an isomorphismU
1
Hom
R
(R/P, M)
Hom
U
1
R
(U
1
R/P, U
1
M), so there is a diagram as above and R/P M is
injective. It follows that P Ass(M).
9.3.11. Proposition. Let R be a noetherian ring and M a module. Then any min-
imal prime P Supp(M) is contained in Ass(M).
Proof. Assume P Supp(M) is minimal. Then the R
P
-module M
P
has support
exactly in the maximal ideal, so PR
P
= Ass(M
P
). Conclusion by 9.3.10.
9.3.12. Denition. A non minimal prime ideal in Ass(M) is an embedded prime
of M.
9.3.13. Theorem. Let R be a noetherian ring and M a nite module. Then the
associated primes Ass(M) is a nite set.
Proof. Follows immediately from 9.3.4 and 8.1.6. Let 0 = M
0
M
n
= M
be a ltration with factors M
i
/M
i1
R/P
i
. Then Ass(M) P
i
.
9.3.14. Corollary. Let R be a noetherian ring and M a nite module. The follow-
ing conditions are equivalent:
(1) M has nite length.
9.3. ASS OF MODULES 131
(2) Supp(M) consists of maximal ideals.
(3) Ass(M) consists of maximal ideals.
(4) Ass(M) = Supp(M).
Proof. Most is just a restatement. (4) (2): Use that the minimal ideals in the
support are maximal.
9.3.15. Lemma. Let (R, P) be a local ring M a module. Then P Ass(M) if
and only if Hom
R
(k(P), M) ,= 0.
Proof. k(P) is simple, so a homomorphism k(P) M is either 0 or injective.
9.3.16. Proposition. Let R be a noetherian ring and M a module. For a prime P
the following conditions are equivalent:
(1) P Ass(M).
(2) PR
P
Ass(M
P
).
(3) Hom
R
P
(k(P), M
P
) ,= 0.
Proof. (1) (2): 9.3.10. (2) (3): 9.3.15.
9.3.17. Theorem. Let R be a noetherian ring and M a nite module. For any
module N
Ass(Hom
R
(M, N)) = Supp(M) Ass(N)
Proof. By 8.2.9 Hom
R
(M, N)
P
Hom
R
P
(M
P
, N
P
). So reduce to the case
where (R,P) is local. Now
Hom
R
(k(P), Hom
R
(M, N)) = Hom
R
(M, Hom
R
(k(P), N))
= Hom
k(P)
(M
R
k(P), Hom
R
(k(P), N))
Conclusion by Nakayamas lemma 6.4.1 and 9.3.15.
9.3.18. Corollary. Let R be a noetherian ring and M a nite module. For a prime
P the following conditions are equivalent:
(1) P Supp(M).
(2) P Ass(Hom
R
(M, R/P)).
9.3.19. Proposition. Let R be a noetherian ring and F a nite module. Assume
rank F
R
k(P) = n for all primes P. Then F is locally free (projective) if and
only if F
P
is free for all P Ass(R).
Proof. Let Qbe a maximal ideal and 0 K R
n
Q
F
Q
0 exact. Ass(K)
Ass(R), so K
P
= 0 for all P Ass(K). By 9.3.7 K = 0.
9.3.20. Theorem. Let (R, P) be a noetherian local ring. If there is a nonzero nite
injective module E, then R is artinian.
Proof. Let Q be a prime and f : R/Q E. If a PQ then a
R/Q
is injective,
so there is f

: R/Q E such that f = f

a
R/Q
. That is P Hom
R
(R/Q, E) =
Hom
R
(R/Q, E), so by Nakayamas lemma 6.4.1 Hom
R
(R/Q, E) = 0 if Q ,= P.
By 9.3.7 0 ,= Hom
R
(R/P, E) Hom
R
(R/Q, E), so Q = P. R is artinian by
8.6.6.
9.3.21. Exercise. (1) Show that
Ass(Z/(n)) = (p)[p prime dividing n
132 9. PRIMARY DECOMPOSITION
(2) Show that
Ass(K[X, Y ]/(X) (X
2
, Y
2
)) = (X), (X, Y )
and point out an embedded prime.
(3) Let I R be an ideal such that

I = I. Show that R/I has no embedded prime


ideals.
(4) Let I, J R be a ideals such that JR
P
IR
P
for all P Ass(R/I). Show that
J I.
9.4. Primary modules
9.4.1. Denition. A submodule N M is a primary submodule or more pre-
cisely P-primary if Ass(M/N) = P. An ideal is a primary ideal if it is a
primary submodule of the ring.
9.4.2. Proposition. A prime ideal P R is a P-primary ideal.
Proof. By 9.3.2 Ass(R/P) = P.
9.4.3. Proposition. Let R be a noetherian ring and M a nite module. For a
proper submodule N M the following conditions are equivalent:
(1) N M is primary for some prime
(2) The set of zero divisors on M/N is contained in the radical
_
Ann(M/N).
(3) For any zero divisor a on M/N there is a power a
n
Ann(M/N).
Proof. (1) (2),(3): Supp(M/N) = V (P) by 9.3.11. So by 9.1.6
_
Ann(M/N) =
P is the set of zero divisors by 9.3.9. (2),(3) (1): By 9.3.7 Ass(M/N) ,= . If
P
1
, P
2
Ass(M/N) then by 9.3.9 the zero divisors in P
1
P
2

_
Ann(M/N)
P
1
P
2
, so P
1
= P
2
.
9.4.4. Corollary. Let R be a noetherian ring and I R a proper ideal. The
following conditions are equivalent:
(1) I R is primary for some prime.
(2) Any zero divisor in the ring R/I is nilpotent.
(3) If ab I for some b / I then some power a
n
I.
9.4.5. Corollary. Let R be a noetherian ring.
(1) If an ideal I R is P-primary then

I = P.
(2) If the radical

I = P is a maximal ideal, then I R is P-primary.


(3) A nite power P
n
R of a maximal ideal is P-primary.
Proof. (1) By 9.3.11 Supp(R/I) = V (P), so

I = P. (2) SuppR/I = V (I) =


P. By 9.3.2 and 9.3.7 Ass(R/I) = P. (3)

P
n
= P.
9.4.6. Proposition. Let R be a noetherian ring and M a module.
(1) If N, N

M are P-primary, then N N

is P-primary.
(2) If N M is P-primary and M/N is nite, then Ann(M/N) R is P-
primary and
_
Ann(M/N) = P.
Proof. (1) From 9.3.5 Ass(M/N N

) Ass(M/N) Ass(M/N

). (2) Let
x
1
, . . . , x
n
be nonzero generators of M/N, then R/ Ann(x
i
) Rx
i
M/N
shows that Ann(x
i
) is P-primary. By (1) Ann(M/N) =
i
Ann(x
i
) is P-primary.
_
Ann(M/N) = P by 9.4.5.
9.4. PRIMARY MODULES 133
9.4.7. Denition. If i : M U
1
M is the canonical homomorphism to the
module of fractions and N

U
1
M is a submodule, then abuse the notation
N

M = i
1
(N

).
9.4.8. Theorem. Let R be a noetherian ring and M a module. Suppose U R is
a multiplicative subset and P is a prime ideal.
(1) If U P ,= and N M is P-primary, then U
1
N = U
1
M.
(2) If U P = and N M is P-primary, then U
1
N U
1
M is PU
1
R-
primary and N = U
1
N M.
(3) If N

U
1
M is PU
1
R-primary, then N

M M is P-primary and
N

= U
1
(N

M).
Proof. From 9.3.10: (1) Ass(U
1
M/N) = . (2) Ass(U
1
M/N) = PR
P
.
Now Ass(U
1
N M/N) Ass(M/N) = P. But (U
1
N M/N)
P
=
N
P
M
P
/N
P
= 0, so N = U
1
N M by 9.3.7. (3) PU
1
R R = P
Ass(M/N

M) is the only associated prime disjoint from U. Let N

M
N M be such that N/N

M R/Q. If Q U ,= , then U
1
N = N

and
therefore N = N

M is a contradiction, so N

M M is P-primary.
9.4.9. Corollary. Let R be a noetherian ring. Suppose U R is a multiplicative
subset and P is a prime ideal.
(1) If U P ,= and I R is P-primary, then IU
1
R = U
1
R.
(2) If U P = and I R is P-primary, then IU
1
R U
1
R is PU
1
R-
primary and I = IU
1
R R.
(3) If I

U
1
R is PU
1
R-primary, then I

R R is P-primary and I

=
(I

R)U
1
R.
9.4.10. Corollary. Let R be a noetherian ring and P a prime ideal.
(1) If P Supp(M) of a nite module, then P
n
M
P
M M is a P-primary
submodule.
(2) P
n
R
P
R R is a P-primary ideal.
Proof. (1) By Nakayamas lemma 6.4.1 Supp(M
P
/P
n
M
P
) = PR
P
. Then
P
n
M
P
M
P
is PR
P
-primary. Conclusion by 9.4.8.
9.4.11. Example. Let R = Z[X] and P = (p, X) for some prime number p. Then
R/P Z/(p), so P is a maximal ideal. The ideal Q = (p
2
, X) satises strict
inclusions P
2
Q P. So

Q = P gives that Q is P-primary, but Q is not a


power of P.
9.4.12. Example. Dene the subring R = Z[pX, X
2
] Z[X] for some prime
number p.
(1) R =

a
i
X[p[a
2i+1
.
(2) The ideal P = (pX, X
2
) =

a
i
X[p[a
2i+1
, a
0
= 0 has R/P Z, so P
is a non maximal prime ideal.
(3) The ideal Q = (q, pX, X
2
) =

a
i
X[p[a
2i+1
, q[a
0
has R/Q Z/(q), so
for q a prime number P Q is a maximal ideal.
(4) The ideal P
2
= (p
2
X
2
, pX
3
, X
4
) =

a
i
X[p[a
2i+1
, a
0
= a
1
= 0, p
2
[a
2

has p pX
2
= p
2
X
2
P
2
, pX
2
/ P
2
and p
n
/ P
2
, so by 9.4.4 P
2
is not
primary.
9.4.13. Exercise. (1) Let K be a eld. Show that (X
2
, Y ) K[X, Y ] is (X, Y )-
primary.
134 9. PRIMARY DECOMPOSITION
(2) Let p be a prime number. Show that (p
k
) Z is a primary ideal.
9.5. Decomposition of modules
9.5.1. Denition. A submodule L M has a primary decomposition if there
exist a family N
i
M of P
i
-primary submodules, such that
L = N
1
N
k
A primary decomposition is a reduced primary decomposition if P
i
,= P
j
for i ,= j
and no N
i
can be excluded.
9.5.2. Lemma. Let R be a noetherian ring and L M submodule with a primary
decomposition. Then by intersection there is a reduced primary decomposition.
Proof. By 9.4.6 intersection of P-primary submodules is P-primary, so replace
more P-primary factors by their intersection.
9.5.3. Lemma. Let R be a noetherian ring and M a nite module.
(1) For each P
i
Ass(M) there is a submodule N
i
M such that Ass(N
i
) =
Ass(M) P
i
and Ass(M/N
i
)) = P
i
.
(2) The intersection

i
N
i
= 0
(3) The module M injects
0 M

i
M/N
i
Proof. The submodule N
i
is given by 9.3.6. Ass(N
i
) = , so conclusion by
9.3.7.
9.5.4. Theorem. Let R be a noetherian ring and M a nite module. A proper
submodule L M has a reduced primary decomposition
L = N
1
N
k
with N
i
P
i
-primary. For any such
(1) The primes are determined
Ass(M/L) = P
1
, . . . , P
k

(2) There is an inclusion


0 M/L

i
M/N
i
Proof. Apply 9.5.3 to M/L to get a reduced primary decomposition. By (2)
Ass(M/L) P
1
, . . . , P
k
. For i there is an inclusion
j=i
N
j
/L
j=i
N
j
+
N
i
/N
i
M/N
i
. It follows since the decomposition is reduced that P
i

Ass(
j=i
N
j
/L) Ass(M/L). So P
1
, . . . , P
k
Ass(M/L).
9.5.5. Proposition. Let R be a noetherian ring and M a nite module. If
L = N
1
N
k
is a reduced primary decomposition of L M and P
i
is one of the minimal primes
in Ass(M/L), then
N
i
= L
P
i
M
9.5. DECOMPOSITION OF MODULES 135
and therefore uniquely determined.
Proof. Since P
i
is minimal, N
j
P
i
= M
P
i
for j ,= i. By 9.4.8 follows that N
i
=
N
P
i
M = L
P
i
M.
9.5.6. Proposition. Let Rbe a noetherian ring and M a nite module. Let L M
such that M/L ,= 0 has nite length. If Ass(M/L) = P
1
, . . . , P
k
, then there is
a reduced primary decomposition
L = N
1
N
k
where
N
i
= L
P
i
M
and an isomorphism
M/L

i
M/N
i
Proof. This follows from 9.5.4, 9.5.5 and 9.2.6. Since M/N
i
have nite length
Supp(M/N
i
) = P
i
and the conditions in 9.2.6 are satised.
9.5.7. Proposition. Let R be a noetherian ring and M a nite length module. If
Ass(M) = P
1
, . . . , P
k
, then there is a reduced primary decomposition
0 = P
1
n
M P
k
n
M
and exact sequences
0 P
n
i
M M M
P
i
0
There are isomorphisms
M

i
M
P
i

i
M/P
i
n
M
Proof. This follows from 7.5.5.
9.5.8. Proposition. Let Rbe a noetherian ring and M a nite module. Let L M
have a reduced primary decomposition
L = N
1
N
k
where N
i
is P
i
-primary. Assume U to be a multiplicative subset disjoint from
exactly P
1
, . . . , P
m
. Then
U
1
L = U
1
N
1
U
1
N
m
is a reduced primary decomposition.
Proof. This follows from 9.4.8 and 9.5.4.
9.5.9. Exercise. (1) Describe the primary decomposition over a eld.
136 9. PRIMARY DECOMPOSITION
9.6. Decomposition of ideals
9.6.1. Theorem. Let R be a noetherian ring. A proper ideal I has a reduced
primary decomposition
I = Q
1
Q
k
with Q
i
P
i
-primary. For any such
(1) The primes are determined
Ass(R/I) = P
1
, . . . , P
k

(2) There is an inclusion


0 R/I

i
R/Q
i
Proof. This is a case of 9.5.4.
9.6.2. Proposition. Let R be a noetherian ring. If
I = Q
1
Q
k
is a reduced primary decomposition and P
i
one of the minimal primes in Ass(R/I),
then
Q
i
= IR
P
i
R
and therefore uniquely determined.
Proof. This is a case of 9.5.5.
9.6.3. Denition. Let P be a prime ideal. The symbolic power of P is
P
(n)
= R P
n
R
P
9.6.4. Lemma. Let P R be a prime ideal in a noetherian ring.
(1) If P is a maximal ideal then P
(n)
= P
n
is P-primary.
(2) P
n
R
P
R
P
is PR
P
-primary.
(3) P
(n)
R is P-primary.
Proof. (1) By 9.4.5 P
n
is primary and by 9.4.9 P
n
= P
(n)
. (2) This follows from
(1). (3) This follws from 9.4.10.
9.6.5. Proposition. Let R be a noetherian ring and let
P
n
= Q
1
Q
k
be a reduced primary decomposition of a power of a prime ideal P. If Q
1
is P-
primary, then
Q
1
= P
(n)
Proof. This is a case of 9.6.2.
9.6.6. Proposition. Let I R be a proper ideal in a noetherian ring such that
R/I is artinian. If Ass(R/I) = P
1
, . . . , P
k
, then there is a reduced primary
decomposition
I = P
(n)
1
P
(n)
k
and an isomorphism
R/I

i
R/P
(n)
i
Proof. This is a case of 9.5.6.
9.6. DECOMPOSITION OF IDEALS 137
9.6.7. Proposition. Let R be an artinian ring. If Ass(R) = P
1
, . . . , P
k
, then
there is a reduced primary decomposition
0 = P
n
1
P
n
k
and exact sequences
0 P
n
i
R R
P
i
0
There are isomorphisms
R

i
R
P
i

i
R/P
n
i
Proof. This is a case of 9.5.7.
9.6.8. Proposition. Let a proper ideal I R in a noetherian ring have a reduced
primary decomposition
I = Q
1
Q
k
where Q
i
is P
i
-primary. Assume U to be a multiplicative subset disjoint from
exactly P
1
, . . . , P
m
. Then
IU
1
R = Q
1
U
1
R Q
m
U
1
R
is a reduced primary decomposition.
Proof. This is a case of 9.5.8.
9.6.9. Proposition. Let R be a noetherian ring and I an ideal. Then

I = P
1
P
k
is the primary decomposition and the elements in Ass(R/

I) = P
1
, . . . , P
k
are
minimal primes over I.
Proof. By 9.3.13 there are only nitely many minimal primes over I. By 5.1.8

I
is the intersection of these and by 9.6.1 this is a reduced primary decomposition
determining the set Ass(R/

I).
9.6.10. Corollary. Let R be a reduced noetherian ring. Then all elements in
Ass(R) are minimal primes. That is, there are no embedded primes.
9.6.11. Corollary. Let Rbe a noetherian ring. The following statements are equiv-
alent:
(1) R is reduced.
(2) R
P
is a eld for all P Ass(R).
(3) R
P
is a domain for all P Ass(R).
Proof. (1) (2): 0 = P
1
P
k
where P
i
are minimal. Then 0 = P
i
R
P
i
is
the decomposition by 9.6.8, so the maximal ideal is zero. (3) (1): Ass(

0)
Ass(R) and

0
P
=

0 R
P
for all P Ass(R). So

0 = 0 by 9.3.7.
9.6.12. Example. Let R be a unique factorization domain. A factorization into
powers of different irreducible primes is a reduced primary decomposition of a
principal ideal.
(1) Let (p) be a prime divisor. The homomorphism 1
p
n1 : R R/(p
n
) ts to
an exact sequence 0 R/(p) R/(p
n
) R/(p
n1
) 0. By 9.3.3 and
induction Ass(R/(p
n
)) Ass(R/(p)) Ass(R/(p
n1
)) = (p) gives that
any power (p
n
) is a (p)-primary ideal.
138 9. PRIMARY DECOMPOSITION
(2) If (a) = (p
n
1
1
) . . . (p
n
k
k
) is a prime factorization then
(a) = (p
n
1
1
) (p
n
k
k
)
0 R/(a) R/(p
n
1
1
) R/(p
n
k
k
)
is a primary decomposition and Ass(R/(a)) = (p
1
), . . . , (p
k
).
9.6.13. Exercise. (1) Let I R be an ideal. Show that if P =

I is a maximal ideal,
then I is a P-primary ideal.
(2) Let I R be an ideal. Show that if I contains a power of a maximal ideal P, then I
is a P-primary ideal.
(3) Let K be a eld and I = (X
2
, XY ) K[X, Y ]. Show that

I = (X), but I is not


(X)-primary.
10
Dedekind rings
10.1. Principal ideal domains
10.1.1. Lemma. Let R be a domain. The set of elements x M in a module with
Ann(x) ,= 0 is a submodule.
Proof. If ax = by = 0 then ab(x +y) = 0. Now use that R is a domain.
10.1.2. Denition. Let Rbe a domain. An element x M in a module is a torsion
element if Ann(x) ,= 0. By 10.1.1 the set of torsion elements is a submodule
T(M) = x M[ Ann(x) ,= 0
of M, the torsion submodule. If T(M) = 0 then M is a torsion free module. If
T(M) = M then M is a torsion module.
10.1.3. Lemma. Let R be a domain and M a module.
(1) M is torsion free if and only if any nonzero a R is a nonzero divisor on M.
(2) The factor M/T(M) is torsion free.
(3) Let K be the fraction eld of R.
T(M) = Ker(M M
(0)
) = Ker(M M
R
K)
(4) If U R is multiplicative, then U
1
T(M) = T(U
1
M).
(5) T is a left exact functor: if 0 M N L is exact, then 0 T(M)
T(N) T(L) is exact.
Proof. (1) x T(M) Ann(x) ,= 0. (2) Let x + T(M) M/T(M) and
0 ,= a R such that ax T(M). Then bax = 0 for some b ,= 0. It follows that
x + T(M) = 0. (3) From 4.4.1 M M
R
K is the canonical homomorphism
M M
(0)
to the fractions by all nonzero denominators, so the kernel is T(M) by
4.2.7. (4) By (3) U
1
T(M) = Ker(U
1
M U
1
M K) = T(U
1
M). (5)
If f : M N is injective and f(x) T(N), then x T(M), so f(T(M)) =
Ker(T(N) T(L)).
10.1.4. Corollary. Let R be a domain and F a at module. Then F is torsion free.
Proof. If 0 ,= a R and F at, then a
F
= a
R
1
F
is injective.
10.1.5. Corollary. Let R be a domain and M a module. The following conditions
are equivalent:
(1) M is torsion free.
(2) M
P
is torsion free for all prime ideals P.
(3) M
P
is torsion free for all maximal ideals P.
Proof. By 10.1.3 T(M)
P
= T(M
P
). Conclusion by the local-global principle
5.4.1.
139
140 10. DEDEKIND RINGS
10.1.6. Corollary. Let R be a noetherian domain and M a module. The following
conditions are equivalent:
(1) M is a torsion module.
(2) (0) / Supp(M).
(3) (0) / Ass(M).
Proof. (1) (2): By 10.1.3 T(M) = Ker(M M
(0)
). (2) (3): By 9.3.11 the
set of primes Ass(M) and Supp(M) have the same minimal elements.
10.1.7. Proposition. Let R be a principal ideal domain. A submodule of a nite
free module is free.
Proof. Let F R
n
be a submodule. If n = 1 F is a principal ideal and free.
Let n > 1 and p : R
n
R be the last projection. Then p(F) is a principal ideal
and free. By induction F Ker p R
n1
is free, so the split exact sequence
0 F Ker p F p(F) 0 gives F F Ker p p(F) is free.
10.1.8. Proposition. Let R be a principal ideal domain.
(1) A torsion free module is at.
(2) A nite torsion free module is free.
(3) A nite torsion module has nite length.
Proof. (1) For a nonzero ideal (a) R the composite R (a) R is a
M
.
For a torsion free F the homomorphism a
F
: F (a)
R
F F is injective.
So F is at by 3.7.12. (2) By (1) a nite torsion free module F is at. By 8.5.7
F is projective and therefore isomorphic to a submodule of a nite free module.
Conclusion by 10.1.7. (3) By 10.1.6 there are only maximal ideals in the support.
Conclusion by 9.2.13.
10.1.9. Proposition. Let R be a principal ideal domain. A nite module M de-
composes
M = T(M) F
as a direct sum of the torsion submodule and a nite free submodule F.
Proof. By 10.1.8 M/T(M) is free, so 0 T(M) M M/T(M) 0 is
split exact. This gives M/T(M) F M.
10.1.10. Proposition. Let R be a principal ideal domain. A nite torsion module
M has a decomposition
M =

(p)Ass(M)
M
p
where (p) is a prime divisor and
M
p
= x M[p
n
x = 0, for some n
Proof. This comes from the primary decomposition 9.5.7. M
p
M/(p
n
)M.
If x M
p
and q ,= p then R = (p, q
n
) Ann(x + (q
n
)M). Therefore M
p
=
Ker(M
q=p
M/(q
n
)M) M/(p
n
)M.
10.1.11. Proposition. Let R be a principal ideal domain and (p) a prime divisor.
A nite torsion module M such that M = M
p
has decomposition
M = R/(p
n
1
) R/(p
n
k
)
where n
1
n
k
.
10.2. DISCRETE VALUATION RINGS 141
Proof. Let n
1
= n be such that p
n
Ann(M), but p
n1
x ,= 0 for some x M.
The short exact sequence 0 Rx M M/Rx 0 of R/(p
n
)-modules
is split exact, since R/(p
n
) Rx is an injective module 7.6.11. Conclusion by
induction on
R
(M).
10.1.12. Theorem. Let R be a principal ideal domain and M a nite module.
Then M has decomposition
M = R/(p
n
1
1
) R/(p
n
k
k
) R
n
where (p
1
), . . . , (p
k
) are not necessarily distinct prime divisors.
10.1.13. Example. (1) A nite abelian group is isomorphic to
Z/(p
n
1
1
) Z/(p
n
k
k
)
where p
1
, . . . , p
k
are not necessarily distinct prime numbers.
(2) A nite dimensional C vector space with a linear map to itself is by 2.1.13 a
C[X] module
C[X]/(X
1
)
n
1
C[X]/(X
k
)
n
k
where
i
are the not necessarily distinct eigenvalues. Choosing C-basis 1, (X
), . . . , (X )
n1
in C[X]/(X )
n
, then the matrix of multiplication by
X is a lower Jordan block
0
B
B
B
@

1
.
.
.
1
1
C
C
C
A
Just use the formulas X(X )
i
= (X )
i+1
+(X )
i
. All together
this gives the Jordan block decomposition.
10.1.14. Exercise. (1) Show that a nonzero nite Z-submodule of Qis a free module
of rank 1.
(2) Show that a nite torsion Z-module is a nite group.
(3) Let K be a eld. Show that a nite torsion K[X]-module is a nite K vector space.
(4) Let R be a noetherian domain such that every nonzero prime ideal is maximal. Let
M be a nite module. Show that the torsion submodule T(M) has nite length.
10.2. Discrete valuation rings
10.2.1. Denition. A local principal ideal domain, which is not a eld, is a dis-
crete valuation ring. A generator of the maximal ideal is a local parameter or a
uniformizing parameter.
10.2.2. Proposition. Let (R, (p)) be a discrete valuation ring. Then

n
(p
n
) = 0
Proof. See Krulls intersection theorem 8.5.5.
10.2.3. Proposition. Let (R, P) be a noetherian local domain. The following
statements are equivalent:
(1) R is a discrete valuation ring.
(2) P is a principal ideal.
(3)
R
(P/P
2
) = 1.
142 10. DEDEKIND RINGS
Proof. (1) (2) (3): Nakayamas lemma 6.4.1. (2) (1): Let P = (p) and
0 ,= a R, by 10.2.2 there is n such that a (p
n
) (p
n+1
). Since any ideal is
nitely generated it follows that a nonzero ideal is of the form (p
n
).
10.2.4. Proposition. Let (R, (p)) be a discrete valuation ring. Any nonzero ideal
is of the form (p
n
) for a unique n = 0, 1, 2, . . . .
Proof. (p) is the only prime divisor. Conclusion by unique factorization 1.5.6.
10.2.5. Corollary. Let (R, (p)) be a discrete valuation ring. Any nonzero element
in the fraction eld K of R has a unique representation up
n
where u is a unit and
n Z.
10.2.6. Denition. Let K be a eld. A surjective map v : K0 Z satisfying
(1) v(xy) = v(x) +v(y)).
(2) If x +y ,= 0, then v(x +y) min(v(x), v(y)).
is a valuation on K.
10.2.7. Proposition. Let v be a valuation on a eld K.
(1) The subset R = x[v(x) 0 0 is subring a of K.
(2) The units in R are x[v(x) = 0.
(3) Any element p, v(p) = 1 generates the same prime divisor (p).
(4) (R, (p)) is a discrete valuation ring.
Proof. (1) Calculate v(1) = 0, so 1 R. If v(x), v(y) 0 then v(xy) 0
and v(x + y) 0, so x, y R xy, x y R. Therefore R is a subring. (2)
v(x) = 0 v(x
1
) = 0. (3), (4) If v(p) = v(q) = 1 then pq
1
is a unit, so
(p) = (q). If 0 ,= x R with v(x) = n > 0 then x = (xp
1
)p (p), so (p) is the
unique maximal ideal.
10.2.8. Denition. The ring in 10.2.7 is the discrete valuation ring of v.
10.2.9. Proposition. Let (R, (p)) be a discrete valuation ring with fraction eld
K. The map
v : K0 Z, up
n
n
is a valuation and R is the discrete valuation ring of v.
Proof. By 10.2.4 v is well dened. Calculate up
m
vp
n
= uvp
m+n
and if m n,
up
m
+vp
n
= (u+vp
nm
)p
m
to see that v is a valuation. Clearly Ris the valuation
ring of v.
10.2.10. Proposition. Let (R, (p)) be a discrete valuation ring. A nite module M
has decomposition
M = R/(p
n
1
) R/(p
n
k
) R
n
where n
1
n
k
> 0.
Proof. This is a case of 10.1.12.
10.2.11. Example. Let (k[[X]], (X)) be the power series ring over a eld. A nite
module is isomorphic to
k[[X]]/(X
n
1
) k[[X]]/(X
n
k
) k[[X]]
n
where n
1
n
k
> 0.
10.3. DEDEKIND DOMAINS 143
10.2.12. Exercise. (1) Let K be a eld. Show that the subring K[[X
2
, X
3
]] K[[X]]
is not a discrete valuation ring.
10.3. Dedekind domains
10.3.1. Denition. A noetherian domain R, which is not a eld, is a Dedekind
domain if all local rings R
P
at nonzero prime ideals are discrete valuation rings.
10.3.2. Proposition. Let R be a Dedekind domain.
(1) Any nonzero prime ideal is maximal.
(2) If U R is multiplicative, then U
1
R is a eld or a Dedekind domain.
Proof. (1) Prime ideals in a nonzero prime ideal P correspond to prime ideals in
the discrete valuation ring R
P
. So (0), P are the only primes and P is maximal.
(2) This is clear from 5.2.13.
10.3.3. Proposition. Let R be a noetherian domain which is not a eld. The fol-
lowing conditions are equivalent:
(1) R is a Dedekind domain.
(2) Every nonzero proper ideal in R is a product of nitely many maximal ideals.
(3)
R
(P/P
2
) = 1 for all maximal ideals.
Proof. (1) (2): By 9.6.4 the primary ideals P
(n)
= P
n
. Conclusion by 9.6.5
and Chinese remainders 1.4.2. (2) (1): Assume (R, P) is local. By Nakayamas
lemma 6.4.1 there is p PP
2
. Since (p) = P
n
it follows that (p) = P and
R is a discrete valuation ring. (1) (3): 10.2.3 and the isomorphism P/P
2

PR
P
/P
2
R
P
.
10.3.4. Proposition. Let R be Dedekind domain.
(1) If R is a unique factorization domain then it is a principal ideal domain.
(2) If R has only nitely many maximal ideals then it is a principal ideal domain.
Proof. (1) Any nonzero prime ideal is principal. Conclusion by 10.3.3. (2) Let
P, P
2
. . . , P
k
be the nitely many maximal ideals. Choose a PP
2
P
2
P
k
,
5.1.3. Then (a) is P primary. By 10.3.3 (a) = P
n
, so (a) = P. As all maximal
ideals are principal, conclusion by 10.3.3.
10.3.5. Proposition. Let R be Dedekind domain. An ideal I is generated by at
most two elements.
Proof. Let Ass(R/I) = P
1
, . . . , P
k
and U = RP
1
P
k
, then by 10.3.4
U
1
R is a principal ideal domain. By 10.1.6 U
1
R U
1
I, so choose by 8.2.9
a homomorphism f : R I such that U
1
f is an isomorphism. Then f is
injective and the ideal f(R) = (a) I satises: P
i
/ Supp(I/(a)) for any i.
Therefore Ass(I/(a)) V (I) = . Let Q
1
, . . . , Q
m
Ass(I/(a)) and choose
b IQ
1
Q
m
. By 9.3.9 b is a nonzero divisor on I/(a) and therefore b
I/(a)
is an isomorphism as I/(a) has nite length, 7.2.6. It follows that I = (a, b).
10.3.6. Theorem. Let R be Dedekind domain.
(1) A torsion free module is at.
(2) A nite torsion free module is projective.
(3) Any ideal is projective.
144 10. DEDEKIND RINGS
(4) Let F be a nite torsion free module. Then there is a number n and an ideal
I such that
F R
n
I
(5) A nite torsion module has nite length.
Proof. (1), (2), (3), (5) These follow from 10.1.6, 6.5.16. (4) Let R have fraction
eld K and assume rank
K
F
R
K = n + 1. Choose a nonzero homomorphism
F R and get by induction on n, F I
1
I
n+1
for nonzero ideals I
j
in R. It sufces to treat the case n = 1. Let Ass(R/I
1
) = P
1
, . . . , P
k
and
U = RP
1
P
k
, then by 10.3.4 U
1
R is a principal ideal domain. By 10.1.8
U
1
I
2
U
1
R, so choose by 8.2.9 a homomorphism f : I
2
R such that
U
1
f is an isomorphism. Then f is injective and the ideal f(I
2
) R satises:
P
i
/ Ass(R/f(I
2
)) for any i. It follows that I
1
+ f(I
2
) = R. Conclusion by a
split exact sequence
0 I I
1
I
2
R 0
10.3.7. Proposition. Let Rbe a Dedekind domain. A nite module M decomposes
M = T(M) F
as a direct sum of the torsion submodule and a nite torsion free submodule F.
Proof. By 10.3.6 the projection M M/T(M) splits.
10.3.8. Proposition. Let R be a Dedekind domain. A nite torsion module M has
a decomposition
M R/P
n
1
1
R/P
n
k
k
where P
1
, . . . , P
k
are not necessarily distinct maximal ideals.
Proof. From 9.5.7 M
P
M
P
and by 10.2.10 the R
P
torsion module M
P

i
R
P
/P
n
i
R
P

i
R/P
n
i
.
10.3.9. Theorem. Let R be a Dedekind domain and M a nite module. Then M
has a decomposition
M = R/P
n
1
1
R/P
n
k
k
R
n
Q
1
Q
l
where P
1
, . . . , P
k
, Q
1
, . . . , Q
l
are not necessarily distinct maximal ideals.
10.3.10. Example. R = Z[X]/(X
2
5) is a noetherian domain. The ideal (2, X+
1) is a prime ideal in Z[X] and X
2
5 = (X+1)
2
2(X+1)4 (2, X+1), so
this corresponds to a prime ideal P R. Calculate P/P
2
= (2, X+1)/(4, 2(X+
1), (X+1)
2
, X
2
5) = (2, X+1)/(4, 2(X+1), (X+1)
2
) which has a nontrivial
submodule (2, (X+1)
2
)/(4, 2(X+1), (X+1)
2
). P/P
2
is not simple, so by 10.3.3
R is not a Dedekind domain.
10.3.11. Exercise. (1) Show that the ring Z[

5] is a Dedekind domain.
(2) Show that the ring Z[

5] is not a Dedekind domain.


Bibliography
A. Altman and S. Kleiman, Introduction to Grothendieck duality theory, Springer-Verlag 1970.
M. Atiyah and I. Macdonald, An introduction to commutative algebra, Addison-Wesley 1969.
N. Bourbaki, Algbre, Hermann 1942-.
N. Bourbaki, Algbre commutative, Hermann-Masson 1961-.
M. Brodmann and R. Sharp, Local cohomology: an algebraic introduction with geometric applica-
tions, Cambridge University Press 1997.
W. Bruns and J. Herzog, Cohen-Macaulay rings, Cambridge University Press 1993.
H. Cartan and S. Eilenberg, Homological algebra, Princeton 1956.
J. Dieudonn, Cours de gomtrie algbrique, Presses Universitaires de France 1974.
D. Eisenbud, Commutative algebra with a view toward algebraic geometry, Springer-Verlag 1996.
E.G. Evans and P. Grifth, Syzygies, Cambridge University Press 1985.
R. Fossum, The divisor class group of a Krull domain, Springer-Verlag 1973.
W. Fulton, Algebraic curves, Benjamin inc. 1969.
W. Fulton, Intercection theory, Springer-Verlag 1984.
R. Gilmer, Multiplicative ideal theory, Marcel Dekker 1972.
A. Grothendieck, Sur quelques points dalggre homologique, Thoku Math, Journ., 9, 1957.
A. Grothendieck, Elments de gomtrie algbrique, IHES 1960-67.
A. Grothendieck, Sminaire de gomtrie algbrique, IHES 1960-67.
A. Grothendieck, Local cohomology, Springer-Verlag 1967.
R. Hartshorne, Residues and duality, Springer-Verlag 1966.
R. Hartshorne, Algebraic geometry, Springer-Verlag 1977.
M. Hochster, Topics in homological theory of modules over commutative rings, American Mathe-
matical Society 1974.
H.C. Hutchins, Examples of commutative rings, Polygonal Publishing House 1981.
B. Iversen, Generic local structure in commutative algebra, Springer-Verlag 1974.
B. Iversen, Cohomology of sheaves, Springer-Verlag 1986.
B. Iversen, Local rings, Aarhus 1974-80.
I. Kaplansky, Commutative rings, Allyn and Bacon 1970
E. Kunz, Introduction to commutative algebra, Birkhuser 1980.
J.P. Lafon, Algbre commutative, Hermann 1977.
S. Lang, Algebra, Addison-Wesley 1965.
H. Matsumura, Commutative algebra, Benjamin inc. 1970.
J. Milne, http://www.jmilne.org.
D. Mumford, Introduction to algebraic geometry, Red and 1 inch thick.
D. Mumford, Algebraic geometry I, complex projective varieties, Springer-Verlag 1976.
M. Nagata, Local rings, Interscience Publ. 1962.
D.G. Northcott, An introduction to homological algebra, Cambridge University Press 1960.
D.G. Northcott, Lessons on rings, modules and multiplicities, Cambridge University Press 1968.
C. Peskine, An algebraic introduction to complex projective geometry, Cambridge University Press
1997.
C. Peskine and L. Szpiro, Dimension projective nie et cohomologie locale, IHES 1973.
M. Reid, Undergraduate commutative algebra, Cambridge University Press 1995.
P.C. Roberts, Homological invariants of modules over commutative rings, Les Presses de
lUniversit de Montral 1980.
P.C. Roberts, Multiplicities and Chern classes in local algebra, Cambridge University Press 1998.
J-P. Serre, Algbre locale - multiplicits, Springer-Verlag 1965.
J-P. Serre, Faisceaux algbrique cohrents, Ann. of Math., 61, 1955.
145
146 BIBLIOGRAPHY
I. Shafarevich, Basic algebraic geometry, Springer-Verlag 1974.
R. Sharp, Steps in commutative algebra, Cambridge University Press 1990.
W.V. Vasconcelos, Computational methods in commutative algebra and algebraic geometry,
Springer-Verlag 1998.
B.L. van der Waerden, Algebra, Springer-Verlag 1967.
O. Zariski and P. Samuel, Commutative algebra, van Nostran 1958-60.
Index
(B), 11
(R, P), 75
(a
ij
), 89
(x
i
), 89
1
M
, 21
1
x
, 27
A/B, 9
A
ij
, 89
I : J, 11
IJ, 11
IM, 23
IS, 11
J R, 11
M(P), 77
M/N, 24
M[
1
u
], 67
M

, 33
M
P
, 77
N

M, 133
N : L, 27
P-primary, 132
P
(n)
, 136
R/I, 11
RY , 23
R[B], 17
R[X], 16
R[[X]], 20
R[
1
u
], 64
R
P
, 75
T(M), 139
U
1
M, 65
U
1
R, 63
U
1
f, 66
V (B), 123
X
B
, 123
Ann, 27
Ass, 128
Cok, 25
Hom
R
, 30
Im, 25
Ker, 11, 25
Spec, 123
Supp, 126
det, 89
C, 10
F
p
, 19
N, 10
Q, 10
R, 10
Z, 10
, 63

R
, 102
, 28
, 35

R
, 34
Q
, 28
rank
R
, 91

I, 14
a
M
, 22
b|a, 15
c(f), 71
char, 12
e

, 29
e
i
, 87
ev
x
, 33
f

, 18
f(P), 77
f

, 33
f
P
, 77
k(P), 75
o(f), 20
v :, 142
a
, 123
0-sequence, 41
abelian group, 9
addition, 9
alternating:, 90
annihilator, 27
artinian module, 105
artinian modules, 105
artinian ring, 107
artinian rings, 107
ass of modules, 128
associated map, 123
associated prime ideal, 128
associative, 9, 21
basis, 29
147
148 INDEX
bilinear, 21
binomial formula, 10
canonical homomorphism, 65
canonical ring homomorphism, 63
Cayley-Hamiltons theorem, 91
change of ring, 37
change of rings, 37
characteristic, 12
Chinese remainder theorem, 14
Chinese remainders, 14
coefcient, 16, 20
cofactor matrix, 89
cokernel, 25
colon ideal, 11, 27
comaximal ideals, 14
commutative, 9
composition series, 102
constants, 16
content of polynomial, 71
contracted ideal, 11
contravariant, 31
cosets, 9
Cramers rule, 90
decomposable, 30
decomposition of ideals, 136
decomposition of modules, 134
Dedekind domain, 143
Dedekind domains, 143
degree, 16
derivative, 18
determinant, 89
direct product, 28
direct sum, 28, 29
discrete valuation ring, 141
discrete valuation ring of v, 142
discrete valuation rings, 141
distributive, 9
divisible module, 58
division, 15
domain, 10
dual homomorphism, 33
dual module, 33
embedded prime, 130
essential extension, 59
evaluation, 33
evaluation map, 17
exact functor, 50, 51
exact sequence, 41
exact sequences, 41
exactness of fractions, 67
exactness of hom, 50
exactness of tensor, 53
extended ideal, 11
factor group, 9
factor module, 24
factor ring, 11
faithfully at, 83
faithfully at ring homomorphism, 84
faithfully at ring homomorphisms, 83
eld, 10
eld extension, 19
elds, 19
nite eld extension, 19
nite ideal, 11
nite length, 102
nite module, 87
nite modules, 87
nite presented module, 94
nite presented modules, 94
nite ring extension, 99
nite ring homomorphism, 99
nite ring homomorphisms, 99
nite type ring, 17
nite type rings, 117
nitely generated ring, 17
ve lemma, 48
at module, 60
at modules, 60
at ring homomorphism, 81
at ring homomorphisms, 81
fraction eld, 64
free module, 29
free modules, 89
Frobenius homomorphism, 12
functor, 31
Gauss lemma, 71
generated, 23
going-down, 85
going-up, 100
Gorenstein ring, 112
greatest common divisor, 16
Hilberts basis theorem, 117
homomorphism, 9, 21
homomorphism module, 30
homomorphism modules, 30
homomorphism modules of fractions, 70
ideal, 11
ideal generated by, 11
ideal product, 11
ideals, 11
idempotent, 15
identity, 9, 21
identity isomorphism, 9, 21
image, 25
indecomposable, 30
induced module, 39
injections, 28
injective envelope, 59
injective module, 56
INDEX 149
injective modules, 56
irreducible component, 125
irreducible element, 15
irreducible principal ideal, 15
irreducible space, 125
irreducible subset, 125
isomorphism, 9, 21
Jacobson radical, 77
kernel, 11, 25
kernel and cokernel, 25
Krulls intersection theorem, 117, 120
Krulls theorem, 73
leading coefcient, 16
least common multiple, 16
left exact contravariant functor, 50
left exact functor, 50
length, 102
linear map, 22
local artinian ring, 110
local parameter, 141
local ring, 75
local ring homomorphism, 75
localization, 109
localization of modules, 77
localization of noetherian rings, 120
localization of rings, 75
localized homomorphism, 77
localized module, 77
localized ring, 75
locally free module, 79
maximal ideal, 13
minimal prime, 126
minimal prime ideal, 74
minor, 89
module, 21
module of fractions, 65
modules and homomorphisms, 21
modules of fractions, 65
monic polynomial, 16
monomial, 16
multilinear:, 89
multiplication, 9
multiplication of principal ideals, 15
multiplicative subset, 63
multiplicity, 18
Nakayamas lemma, 93
natural homomorphism, 31
natural isomorphism, 31
negative, 9
nilpotent, 14
nilradical, 14
noetherian module, 113
noetherian ring, 115
noetherian rings, 115
noncommutative ring, 9
nontrivial idempotent, 15
nonzero divisor, 10, 22
normed:, 90
notherian modules, 113
order, 20
polynomial, 16
polynomial ring, 16
polynomials, 16
power series, 20
power series ring, 20
power series rings, 118
primary decomposition, 134
primary ideal, 132
primary modules, 132
primary submodule, 132
prime divisor, 15
prime element, 15
prime elds, 19
prime ltrations of modules, 121
prime ideal, 13
prime ideals, 13, 73
principal ideal, 11
principal ideal domain, 15
principal ideal domains, 139
principal open subsets, 123
product ring, 10
projection, 9, 24
projections, 28
projective module, 55
projective modules, 55
proper ideal, 11
radical, 14
rank, 91
reduced, 14
reduced primary decomposition, 134
reexive module, 33
residue eld, 75
residue homomorphism, 77
restriction of scalars, 22
retraction, 43
right exact contravariant functor , 50
right exact functor, 50
ring, 9
ring extension, 9
ring generated, 17
ring of fractions, 63
rings, 9
rings of fractions, 63
root, 18
roots, 18
scalar multiplication, 21, 22
section, 43
semi-local ring, 128
150 INDEX
short exact sequence, 43
simple module, 101
simple modules, 101
simple root, 18
snake homomorphism, 46
snake lemma, 47
spectrum, 123
split exact sequence, 44
standard basis, 29
subeld, 19
subgroup, 9
submodule, 21
submodule generated, 23
submodules and factor modules, 23
subring, 9
sum and product, 28
support, 126
support of modules, 126
symbolic power, 136
tensor modules of fractions, 69
tensor product, 34
tensor product modules, 34
tensor product ring, 39
the length, 102
The local-global principle, 79
the polynomial ring is factorial, 71
the snake lemma, 45
torsion element, 139
torsion free module, 139
torsion module, 139
torsion submodule, 139
total ring of fractions, 64
uniformizing parameter, 141
unique factorization, 15
unique factorization domain, 15
unit, 10
valuation, 142
vector space, 22
windmill lemma, 49
Zariski topology, 123
zero, 9
zero divisor, 10, 22
zero ideal, 11
zero module, 21
zero submodule, 21