Chapter 2

Ring Theory
In the rst section below, a ring will be dened as an abstract structure with
a commutative addition, and a multiplication which may or may not be com-
mutative. This distinction yields two quite dierent theories: the theory of
respectively commutative or non-commutative rings. These notes are mainly
concerned about commutative rings.
Non-commutative rings have been an object of systematic study only quite
recently, during the 20th century. Commutative rings on the contrary have
appeared though in a hidden way much before, and as many theories, it all goes
back to Fermats Last Theorem.
In 1847, the mathematician Lame announced a solution of Fermats Last
Theorem, but Liouville noticed that the proof depended on a unique decompo-
sition into primes, which he thought was unlikely to be true. Though Cauchy
supported Lame, Kummer was the one who nally published an example in
1844 to show that the uniqueness of prime decompositions failed. Two years
later, he restored the uniqueness by introducing what he called ideal complex
numbers (today, simply ideals) and used it to prove Fermats Last Theorem
for all n < 100 except n = 37, 59, 67 and 74.
It is Dedekind who extracted the important properties of ideal numbers,
dened an ideal by its modern properties: namely that of being a subgroup
which is closed under multiplication by any ring element. He further introduced
prime ideals as a generalization of prime numbers. Note that today we still
use the terminology Dedekind rings to describe rings which have in particu-
lar a good behavior with respect to factorization of prime ideals. In 1882, an
important paper by Dedekind and Weber developed the theory of rings of poly-
nomials. At this stage, both rings of polynomials and rings of numbers (rings
appearing in the context of Fermats Last Theorem, such as what we call now
the Gaussian integers) were being studied. But it was separately, and no one
made connection between these two topics. Dedekind also introduced the term
eld (Korper) for a commutative ring in which every non-zero element has a
57
58 CHAPTER 2. RING THEORY
multiplicative inverse but the word ring is due to Hilbert, who, motivated by
studying invariant theory, studied ideals in polynomial rings proving his famous
Basis Theorem in 1893.
It will take another 30 years and the work of Emmy Noether and Krull to see
the development of axioms for rings. Emmy Noether, about 1921, is the one who
made the important step of bringing the two theories of rings of polynomials
and rings of numbers under a single theory of abstract commutative rings.
In contrast to commutative ring theory, which grew from number theory,
non-commutative ring theory developed from an idea of Hamilton, who at-
tempted to generalize the complex numbers as a two dimensional algebra over
the reals to a three dimensional algebra. Hamilton, who introduced the idea of
a vector space, found inspiration in 1843, when he understood that the gener-
alization was not to three dimensions but to four dimensions and that the price
to pay was to give up the commutativity of multiplication. The quaternion
algebra, as Hamilton called it, launched non-commutative ring theory.
Other natural non-commutative objects that arise are matrices. They were
introduced by Cayley in 1850, together with their laws of addition and multi-
plication and, in 1870, Pierce noted that the now familiar ring axioms held for
square matrices.
An early contributor to the theory of non-commutative rings was the Scottish
mathematician Wedderburn, who in 1905, proved Wedderburns Theorem,
namely that every nite division ring is commutative and so is a eld.
It is only around the 1930s that the theories of commutative and non-
commutative rings came together and that their ideas began to inuence each
other.
2.1 Rings, ideals and homomorphisms
Denition 2.1. A ring R is an abelian group with a multiplication operation
(a, b) ab
which is associative, and satises the distributive laws
a(b +c) = ab +ab, (a +b)c = ac +bc
with identity element 1.
There is a group structure with the addition operation, but not necessarily
with the multiplication operation. Thus an element of a ring may or may not be
invertible with respect to the multiplication operation. Here is the terminology
used.
Denition 2.2. Let a, b be in a ring R. If a ,= 0 and b ,= 0 but ab = 0, then
we say that a and b are zero divisors. If ab = ba = 1, we say that a is a unit or
that a is invertible.
2.1. RINGS, IDEALS AND HOMOMORPHISMS 59
While the addition operation is commutative, it may or not be the case with
the multiplication operation.
Denition 2.3. Let R be ring. If ab = ba for any a, b in R, then R is said to
be commutative.
Here are the denitions of two particular kinds of rings where the multipli-
cation operation behaves well.
Denition 2.4. An integral domain is a commutative ring with no zero divisor.
A division ring or skew eld is a ring in which every non-zero element a has an
inverse a
1
.
Let us give two more denitions and then we will discuss several examples.
Denition 2.5. The characteristic of a ring R, denoted by charR, is the small-
est positive integer such that
n 1 = 1 + 1 +. . . + 1
. .
ntimes
= 0.
We can also extract smaller rings from a given ring.
Denition 2.6. A subring of a ring R is a subset S of R that forms a ring
under the operations of addition and multiplication dened in R.
Examples 2.1. 1. Z is an integral domain but not a eld.
2. The integers modulo n form a ring, which is an integral domain if and
only if n is prime.
3. The n n matrices /
n
(R) with coecients in R are a ring, but not an
integral domain if n 2.
4. Let us construct the smallest and also most famous example of division
ring. Take 1, i, j, k to be basis vectors for a 4-dimensional vector space
over R, and dene multiplication by
i
2
= j
2
= k
2
= 1, ij = k, jk = i, ki = j, ji = ij, kj = jk, ik = ki.
Then
H = a +bi +cj +dk, a, b, c, d R
forms a division ring, called the Hamiltons quaternions. So far, we have
only seen the ring structure. Let us now discuss the fact that every non-
zero element is invertible. Dene the conjugate of an element h = a+bi +
cj +dk H to be

h = a bi cj dk (yes, exactly the same way you did
it for complex numbers). It is an easy computation (and a good exercise
if you are not used to the non-commutative world) to check that
q q = a
2
+b
2
+c
2
+d
2
.
60 CHAPTER 2. RING THEORY
Now take q
1
to be
q
1
=
q
q q
.
Clearly qq
1
= q
1
q = 1 and the denominator cannot possibly be 0, but
if a = b = c = d = 0.
5. If R is a ring, then the set R[X] of polynomials with coecients in R is a
ring.
Similarly to what we did with groups, we now dene a map from a ring to
another which has the property of carrying one ring structure to the other.
Denition 2.7. Let R, S be two rings. A map f : R S satisfying
1. f(a +b) = f(a) +f(b) (this is thus a group homomorphism)
2. f(ab) = f(a)f(b)
3. f(1
R
) = 1
S
for a, b R is called ring homomorphism.
The notion of ideal number was introduced by the mathematician Kum-
mer, as being some special numbers (well, nowadays we call them groups)
having the property of unique factorization, even when considered over more
general rings than Z (a bit of algebraic number theory would be good to make
this more precise). Today only the name ideal is left, and here is what it gives
in modern terminology:
Denition 2.8. Let 1 be a subset of a ring R. Then an additive subgroup of
R having the property that
ra 1 for a 1, r R
is called a left ideal of R. If instead we have
ar 1 for a 1, r R
we say that we have a right ideal of R. If an ideal happens to be both a right
and a left ideal, then we call it a two-sided ideal of R, or simply an ideal of R.
Of course, for any ring R, both R and 0 are ideals. We thus introduce
some terminology to precise whether we consider these two trivial ideals.
Denition 2.9. We say that an ideal 1 of R is proper if 1 , = R. We say that
is it non-trivial if 1 , = R and 1 , = 0.
If f : R S is a ring homomorphism, we dene the kernel of f in the most
natural way:
Kerf = r R, f(r) = 0.
Since a ring homomorphism is in particular a group homomorphism, we already
know that f is injective if and only if Kerf = 0. It is easy to check that Kerf
is a proper two-sided ideal:
2.2. QUOTIENT RINGS 61
Kerf is an additive subgroup of R.
Take a Kerf and r R. Then
f(ra) = f(r)f(a) = 0 and f(ar) = f(a)f(r) = 0
showing that ra and ar are in Kerf.
Then Kerf has to be proper (that is, Kerf ,= R), since f(1) = 1 by
denition.
We can thus deduce the following (extremely useful) result.
Lemma 2.1. Suppose f : R S is a ring homomorphism and the only two-
sided ideals of R are 0 and R. Then f is injective.
Proof. Since Kerf is a two-sided ideal of R, then either Kerf = 0 or Kerf = R.
But Kerf ,= R since f(1) = 1 by denition (in words, Kerf is a proper ideal).
At this point, it may be worth already noticing the analogy between on the
one hand rings and their two-sided ideals, and on the other hand groups and
their normal subgroups.
Two-sided ideals are stable when the ring acts on them by multiplication,
either on the right or on the left, and thus
rar
1
1, a 1, r R,
while normal subgroups are stable when the groups on them by conjuga-
tion
ghg
1
H, h H, g G (H G).
Groups with only trivial normal subgroups are called simple. We will not
see it formally here, but rings with only trivial two-sided ideals as in the
above lemma are called simple rings.
The kernel of a group homomorphism is a normal subgroup, while the
kernel of a ring homomorphism is an ideal.
Normal subgroups allowed us to dene quotient groups. We will see now
that two-sided ideals will allow to dene quotient rings.
2.2 Quotient rings
Let 1 be a proper two-sided ideal of R. Since 1 is an additive subgroup of R
by denition, it makes sense to speak of cosets r +1 of 1, r R. Furthermore,
a ring has a structure of abelian group for addition, so 1 satises the denition
of a normal subgroup. From group theory, we thus know that it makes sense to
speak of the quotient group
R/1 = r +1, r R,
62 CHAPTER 2. RING THEORY
group which is actually abelian (inherited from R being an abelian group for
the addition).
We now endow R/1 with a multiplication operation as follows. Dene
(r +1)(s +1) = rs +1.
Let us make sure that this is well-dened, namely that it does not depend on
the choice of the representative in each coset. Suppose that
r +1 = r

+1, s +1 = s

+1,
so that a = r

r 1 and b = s

s 1. Now
r

= (a +r)(b +s) = ab +as +rb +rs rs +1

since ab, as and rb belongs to 1 using that a, b 1 and the denition of ideal.
This tells us r

is also in the coset rs + 1 and thus multiplication does not

depend on the choice of representatives. Note though that this is true only
because we assumed a two-sided ideal 1, otherwise we could not have concluded,
since we had to deduce that both as and rb are in 1.
Denition 2.10. The set of cosets of the two-sided ideal 1 given by
R/1 = r +1, r R
is a ring with identity 1
R
+1 and zero element 0
R
+1 called a quotient ring.
Note that we need the assumption that 1 is a proper ideal of R to claim that
R/1 contains both an identity and a zero element (if R = 1, then R/1 has only
one element).
Example 2.2. Consider the ring of matrices /
2
(F
2
[i]), where F
2
denotes the
integers modulo 2, and i is such that i
2
= 1 1 mod 2. This is thus the ring
of 2 2 matrices with coecients in
F
2
[i] = a +ib, a, b 0, 1.
Let 1 be the subset of matrices with coecients taking values 0 and 1 +i only.
It is a two-sided ideal of /
2
(F
2
[i]). Indeed, take a matrix U 1, a matrix
M /
2
(F
2
[i]), and compute UM and MU. An immediate computation shows
that all coecients are of the form a(1+i) with a F
2
[i], that is all coecients
are in 0, 1 +i. Clearly 1 is an additive group.
We then have a quotient ring
/
2
(F
2
[i])/1.
We have seen that Kerf is a proper ideal when f is a ring homomorphism.
We now prove the converse.
Proposition 2.2. Every proper ideal 1 is the kernel of a ring homomorphism.
2.2. QUOTIENT RINGS 63
Proof. Consider the canonical projection that we know from group theory.
Namely
: R R/1, r (r) = r +1.
We already know that is group homomorphism, and that its kernel is 1. We
are only left to prove that is a ring homomorphism:
(rs) = rs +1 = (r +1)(s +1) = (r)(s).
(1
R
) = 1
R
+1 which is indeed the identity element of R/1.
We are now ready to state a factor theorem and a 1st isomorphism theorem
for rings, the same way we did for groups. It may help to keep in mind the
analogy between two-sided ideals and normal subgroups mentioned above.
Assume that we have a ring R which contains a proper ideal 1, another ring
S, and f : R S a ring homomorphism. Let be the canonical projection
from R to the quotient group R/1:
R S
R/1
?

-
f

f
We would like to nd a ring homomorphism

f : R/1 S that makes the
diagram commute, namely
f(a) =

f((a))
for all a R.
Theorem 2.3. (Factor Theorem for Rings). Any ring homomorphism f
whose kernel K contains 1 can be factored through R/1. In other words, there
is a unique ring homomorphism

f : R/1 S such that

f = f. Furthermore
1.

f is an epimorphism if and only if f is.
2.

f is a monomorphism if and only if K = 1.
3.

f is an isomorphism if and only if f is an epimorphism and K = 1.
Proof. Since we have already done the proof for groups with many details, here
we will just mention a few important points in the proof.
Let a +1 R/1 such that (a) = a +1 for a R. We dene

f(a +1) = f(a).

This is the most natural way to do it, however, we need to make sure that this
is indeed well-dened, in the sense that it should not depend on the choice of
the representative taken in the coset. Let us thus take another representative,
64 CHAPTER 2. RING THEORY
say b a +1. Since a and b are in the same coset, they satisfy a b 1 K,
where K = Ker(f) by assumption. Since a b K, we have f(a b) = 0 and
thus f(a) = f(b).
Now that

f is well dened, it is an easy computation to check that

f inherits
the property of ring homomorphism from f.
The rest of the proof works exactly the same as for groups.
The rst isomorphism theorem for rings is similar to the one for groups.
Theorem 2.4. (1st Isomorphism Theorem for Rings). If f : R S is a
ring homomorphism with kernel K, then the image of f is isomorphic to R/K:
Im(f) R/Ker(f).
Proof. We know from the Factor Theorem that

f : R/Ker(f) S
is an isomorphism if and only if f is an epimorphism, and clearly f is an epi-
morphism on its image, which concludes the proof.
Example 2.3. Let us nish Example 2.2. We showed there that /
2
(F
2
[i])/1
is a quotient ring, where 1 is the ideal formed of matrices with coecients in
0, 1 +i. Consider the ring homomorphism:
f : /
2
(F
2
[i]) /
2
(F
2
), M = (m
k,l
) f(M) = (m
k,l
mod 1 +i)
that is f looks at the coecients of M mod 1 + i. Its kernel is 1 and it is
surjective. By the rst isomorphism for rings, we have
/
2
(F
2
[i])/1 /
2
(F
2
).
2.3 The Chinese Remainder Theorem
We will prove a general Chinese Remainder Theorem, rephrased in terms of
rings and ideals.
For that let us start by introducing some new denitions about ideals, that
will collect some of the manipulations one can do on ideals. Let us start with
the sum.
Denition 2.11. Let 1 and be two ideals of a ring R. The sum of 1 and
is the ideal
1 + = x +y, x 1, y .
If 1 and are right (resp. left) ideals, so is their sum.
Note that the intersection 1 of two (resp. right, left, two-sided) ideals
of R is again a (resp. right, left, two-sided) ideal of R. We can dene a notion
of being co-prime for ideals as follows.
2.3. THE CHINESE REMAINDER THEOREM 65
Denition 2.12. The ideals 1 and of R a commutative ring are relatively
prime if
1 + = R.
Finally, let us extend the notion of modulo to ideals.
Denition 2.13. If a, b R and 1 is an ideal of R, we say that a is congruent
to b modulo 1 if
a b 1.
A last denition this time about rings is needed before we can state the
theorem.
Denition 2.14. If R
1
, . . . , R
n
are rings, the direct product of the R
i
is dened
as the ring of n-tuples (a
1
, . . . , a
n
), a
i
R
i
, with componentwise addition and
multiplication. The zero element is (0, . . . , 0) and the identity is (1, . . . , 1) where
1 means 1
R
i
for each i.
This denition is an immediate generalization of the direct product we stud-
ied for groups.
Theorem 2.5. (Chinese Remainder Theorem). Let R be a commutative
ring, and let 1
1
, . . . , 1
n
be ideals in R, such that
1
i
+1
j
= R, i ,= j.
1. If a
1
, . . . , a
n
are elements of R, there exists an element a R such that
a a
i
mod 1
i
, i = 1, . . . , n.
2. If b is another element of R such that b a
i
mod 1
i
, i = 1, . . . , n, then
b a mod
n
i=1
1
i
.
Conversely, if b satises the above congruence, then b a
i
mod 1
i
, i =
1, . . . , n.
3. We have that
R/
n
i=1
1
i

n

i=1
R/1
i
.
Proof. 1. For j > 1, we have by assumption that 1
1
+1
j
= R, and thus there
exist b
j
1
1
and d
j
1
j
such that
b
j
+d
j
= 1, j = 2, . . . , n.
This yields that
n

j=2
(b
j
+d
j
) = 1. (2.1)
66 CHAPTER 2. RING THEORY
Now if we look at the left hand side of the above equation, we have
(b
2
+d
2
)(b
3
+d
3
) (b
n
+d
n
) = (b
2
b
3
+b
2
d
3
+d
2
b
3
. .
I
1
+d
2
d
3
) (b
n
+d
n
)
and all the terms actually belong to 1
1
, but c
1
:=

n
j=2
d
j

n
j=2
1
j
.
Thus
c
1
1 mod 1
1
from (2.1). On the other hand, we also have
c
1
0 mod 1
j
for j > 1 since c
1

n
j=2
1
j
.
More generally, for all i, we can nd c
i
with
c
i
1 mod 1
i
, c
i
0 mod 1
j
, j ,= i.
Now take arbitrary elements a
1
, . . . , a
n
R, and set
a = a
1
c
1
+. . . +a
n
c
n
.
Let us check that a is the solution we are looking for. Rewrite
a a
i
= a a
i
c
i
+a
i
c
i
a
i
= (a a
i
c
i
) +a
i
(c
i
1).
If we look modulo 1
i
, we get
aa
i
aa
i
c
i
a
1
+. . . +a
i1
c
i1
+a
i+1
c
i+1
+. . . +a
n
c
n
0 mod 1
i
where the rst congruence follows from c
i
1 0 mod 1
i
and the third
congruence comes from c
j
0 mod 1
j
,j ,= i.
2. We have just shown the existence of a solution a modulo 1
i
for i = 1, . . . , n.
We now discuss the question of unicity, and show that the solution is
actually not unique, but any other solution than a is actually congruent
to a mod
n
i=1
1
i
.
We have for all i = 1, . . . , n that
b a
i
mod 1
i
b a mod 1
i
b a 0 mod 1
i
which nally is equivalent to
b a
n
i=1
1
i
.
3. Dene the ring homomorphism f : R

n
i=1
R/1
i
, sending
a f(a) = (a +1
1
, . . . , a +1
n
).
2.4. MAXIMAL AND PRIME IDEALS 67
This map is surjective: for any (a
1
+ 1
1
, . . . , a
n
+ 1
n
)

n
i=1
R/1
i
,
there exists an a R such that f(a) = (a
1
+ 1
1
, . . . , a
n
+ 1
n
), that
is a
i
a mod 1
i
, by the rst point.
Its kernel is given by
Kerf = a R, f(a) = (1
1
, . . . , 1
n
)
= a R, a 1
i
, i = 1, . . . , n
=
n

i=1
1
i
.
We conclude using the rst isomorphism Theorem for rings.
The Chinese remainder Theorem does not hold in the non-commutative case.
Consider the ring R of non-commutative real polynomials in X and Y . Denote
by 1 the principal two-sided ideal generated by X and the principal two-sided
ideal generated by XY + 1. Then 1 + = R but 1 , = 1.
2.4 Maximal and prime ideals
Here are a few special ideals.
Denition 2.15. The ideal generated by the non-empty set X of R is the
smallest ideal of R that contains X. It is denoted by X). It is the collection
of all nite sums of the form

i
r
i
x
i
s
i
.
Denition 2.16. An ideal generated by a single element a is called a principal
ideal, denoted by a).
Denition 2.17. A maximal ideal in the ring R is a proper ideal that is not
contained in any strictly larger proper ideal.
One can prove that every proper ideal is contained in a maximal ideal, and
that consequently every ring has at least one maximal ideal. We skip the proof
here, since it heavily relies on set theory, requires many new denitions and the
use of Zorns lemma.
Instead, let us mention that a correspondence Theorem exists for rings, the
same way it exists for groups, since we will need it for characterizing maximal
ideals.
Theorem 2.6. (Correspondence Theorem for rings). If 1 is an ideal of
a ring R, then the canonical map
: R R/1
sets up a one-to-one correspondence between
68 CHAPTER 2. RING THEORY
the set of all subrings of R containing 1 and the set of all subrings of R/1,
the set of all ideals of R containing 1 and the set of all ideals of R/1.
Here is a characterization of maximal ideals in commutative rings.
Theorem 2.7. Let M be an ideal in the commutative ring R. We have
M maximal R/M is a eld.
Proof. Let us start by assuming that M is maximal. Since R/M is a ring, we
need to nd the multiplicative inverse of a+M R/M assuming that a+M ,= 0
in R/M, that is for a , M. Since M is maximal, the ideal Ra +M has to be R
itself, since M Ra +M. Thus 1 Ra +M = R, that is
1 = ra +m, r R, m M.
Then
(r +M)(a +M) = ra +M = (1 m) +M = 1 +M
proving that r +M is (a +M)
1
.
Conversely, let us assume that R/M is a eld. First we notice that M must
be a proper ideal of R, since if M = R, then R/M contains only one element
and 1 = 0.
Let N be an ideal of R such that M N R and N ,= R. We have to
prove that M = N to conclude that M is maximal.
By the correspondence Theorem for rings, we have a one-to-one correspon-
dence between the set of ideals of R containing M, and the set of ideals of R/M.
Since N is such an ideal, its image (N) R/M must be an ideal of R/M, and
thus must be either 0 or R/M. The latter yields that N = R, which is a
contradiction, letting as only possibility that (N) = 0, and thus N = M,
which completes the proof.
Denition 2.18. A prime ideal in a commutative ring R is a proper ideal P
of R such that for any a, b R, we have that
ab P a P or b P.
Here is again a characterization of a prime ideal P of R in terms of its
quotient ring R/P.
Theorem 2.8. If P is an ideal in the commutative ring R
P is a prime ideal R/P is an integral domain.
Proof. Let us start by assuming that P is prime. It is thus proper by denition,
and R/P is a ring. We must show that the denition of integral domain holds,
namely that
(a +P)(b +P) = 0 +P a +P = P or b +P = P.
2.5. POLYNOMIAL RINGS 69
Since
(a +P)(b +P) = ab +P = 0 +P,
we must have ab P, and thus since P is prime, either a P or b P, implying
respectively that either a +P = P or b +P = P.
Conversely, if R/P is an integral domain, then P must be proper (otherwise
1 = 0). We now need to check the denition of a prime ideal. Let us thus
consider ab P, implying that
(a +P)(b +P) = ab +P = 0 +P.
Since R/P is an integral domain, either a +P = P or b +P = P, that is
a P or b P,
which concludes the proof.
Corollary 2.9. In a commutative ring, a maximal ideal is prime.
Proof. If M is maximal, then R/M is a eld, and thus an integral domain, so
that M is prime.
Corollary 2.10. Let f : R S be an epimorphism of commutative rings.
1. If S is a eld, then Kerf is a maximal ideal of R.
2. If S is an integral domain, then Kerf is a prime ideal of R.
Proof. By the rst isomorphism theorem for rings, we have that
S R/Kerf.
Example 2.4. Consider the ring Z[X] of polynomials with coecients in Z, and
the ideal generated by the indeterminate X, that is X) is the set of polynomials
with constant coecient 0. Clearly X) is a proper ideal. To show that it is
prime, consider the following ring homomorphism:
: Z[X] Z, f(X) (f(X)) = f(0).
We have that X) = Ker which is prime by the above corollary.
2.5 Polynomial rings
For this section, we assume that R is a commutative ring. Set R[X] to be the
set of polynomials in the indeterminate X with coecients in R. It is easy to
see that R[X] inherits the properties of ring from R.
70 CHAPTER 2. RING THEORY
We dene the evaluation map E
x
, which evaluates a polynomial f(X)
R[X] in x R, as
E
x
: R[X] R, f(X) f(X)[
X=x
= f(x).
We can check that E
x
is a ring homomorphism.
The degree of a polynomial is dened as usual, that is, if p(X) = a
0
+a
1
X+
. . . + a
n
X
n
with a
n
,= 0, then deg(p(X)) = deg p = n. By convention, we set
deg(0) = .
Euclidean division will play an important role in what will follow. Let us
start by noticing that there exists a polynomial division algorithm over R[X],
namely: if f, g R[X], with g monic, then there exist unique polynomials q
and r in R[X] such that
f = qg +r, deg r < deg g.
The requirement that g is monic comes from R being a ring and not necessarily
a eld. If R is a eld, g does not have to be monic, since one can always multiply
g by the inverse of the leading coecient, which is not possible if R is not a
eld.
This gives the following:
Theorem 2.11. (Remainder Theorem). If f R[X], a R, then there
exists a unique polynomial q(X) R[X] such that
f(X) = q(X)(X a) +f(a).
Hence f(a) = 0 X a [ f(X).
Proof. Since (X a) is monic, we can do the division
f(X) = q(X)(X a) +r(X).
But now since deg r < deg(X a), r(X) must be a constant polynomial, which
implies that
f(a) = r(X)
and thus
f(X) = q(X)(X a) +f(a)
as claimed. Furthermore, we clearly have that
f(a) = 0 X a [ f(X).
The following result sounds well known, care should be taken not to gener-
alize it to rings which are not integral domain!
Theorem 2.12. If R is an integral domain, then a non-zero polynomial f in
R[X] of degree n has at most n roots in R, counting multiplicity.
2.6. UNIQUE FACTORIZATION AND EUCLIDEAN DIVISION 71
Proof. Take a
1
a root of f, that is f(a
1
) = 0. Then
X a
1
[ f(X)
by the remainder Theorem above, meaning that
f(X) = q
1
(X)(X a
1
)
n
1
where q
1
(a
1
) ,= 0 and deg q
1
= n n
1
. Similarly, consider a
2
,= a
1
another root
of f, so that
0 = f(a
2
) = q
1
(a
2
)(a
2
a
1
)
n
1
.
Since R is an integral domain, we must have that q
1
(a
2
) = 0, and thus a
2
is a
root of q
1
(X). We can repeat the process with q
1
(X) instead of f(X): since a
2
is a root of q
1
(X), we have
q
1
(X) = q
2
(X)(X a
2
)
n
2
with q
2
(a
2
) ,= 0 and deg q
2
= n n
1
n
2
. By going on iterating the process,
we obtain
f(X) = q
1
(X)(X a
1
)
n
1
= q
2
(X)(X a
2
)
n
2
(X a
1
)
n
1
= . . .
= (X a
1
)
n
1
(X a
2
)
n
2
(X a
k
)
n
k
c
where c is some constant and
n = n
1
+n
2
+ +n
k
.
Since R is an integral domain, the only possible roots of f are a
1
, . . . , a
k
, k
n.
Example 2.5. Take R = Z
8
the ring of integers modulo 8. Consider the
polynomial
f(X) = X
3
.
It is easy to check that is has 4 roots: 0, 2, 4, 6. This comes from the fact that
Z
8
is not an integral domain.
2.6 Unique factorization and Euclidean division
In this section, all rings are assumed to be integral domains.
Let us start by dening formally the notions of irreducible and prime. The
elements a, b, c, u in the denitions below all belong to an integral domain R.
Denition 2.19. The elements a, b are called associate if a = ub for some unit
u.
72 CHAPTER 2. RING THEORY
Denition 2.20. Let a be a non-zero element which is not a unit. Then a is
said to be irreducible if a = bc implies that either b or c must be a unit.
Denition 2.21. Let a be a non-zero element which is not a unit. Then a is
called prime is whenever a [ bc, then a [ b or a [ c.
Between prime and irreducible, which notion is the stronger? The answer is
in the proposition below.
Proposition 2.13. If a is prime, then a is irreducible.
Proof. Suppose that a is prime, and that a = bc. We want to prove that either
b or c is a unit. By denition of prime, we must have that a divides either b or
c. Let us say that a divides b. Thus
b = ad b = bcd b(1 cd) = 0 cd = 1
using that R is an integral domain, and thus c is a unit. The same argument
works if we assume that a divides c, and we conclude that a is irreducible.
Example 2.6. Consider the ring
R = Z[

3] = a +ib

3, a, b Z.
We want to see that 2 is irreducible but not prime.
Let us rst check that 2 is indeed irreducible. Suppose that
2 = (a +ib

3)(c +id

3).
Since 2 is real, it is equal to its conjugate, and thus
2

2 = (a +ib

3)(c +id

3)(a ib

3)(c id

3)
implies that
4 = (a
2
+ 3b
2
)(c
2
+ 3d
2
).
We deduce that a
2
+3b
2
must divide 4, and it cannot possibly be 2, since
we have a sum of squares in Z. If a
2
+ 3b
2
= 4, then c
2
+ 3d
2
= 1 and
d = 0, c = 1. Vice versa if c
2
+ 3d
2
= 4 then a
2
+ 3b
2
= 1, and b = 0,
a = 1. In both cases we get that one of the factors of 2 is unit, namely
1.
We now have to see that 2 is not a prime. Clearly
2 [ (1 +i

3)(1 i

3) = 4.
But 2 divides neither 1 +i

3 nor 1 i

3.
We can see from the above example that the problem which arises is the lack
of unique factorization.
2.6. UNIQUE FACTORIZATION AND EUCLIDEAN DIVISION 73
Denition 2.22. A unique factorization domain (UFD) is an integral domain
R satisfying that
1. every element 0 ,= a R can be written as a product of irreducible factors
p
1
, . . . p
n
up to a unit u, namely:
a = up
1
. . . p
n
.
2. The above factorization is unique, that is, if
a = up
1
. . . p
n
= vq
1
. . . q
m
are two factorizations into irreducible factors p
i
and q
j
with units u, v,
then n = m and p
i
and q
i
are associate for all i.
We now prove that the distinction between irreducible and prime disappear
in a unique factorization domain.
Proposition 2.14. In a unique factorization domain R, we have that a is
irreducible if and only if a is prime.
Proof. We already know that prime implies irreducible. Let us show that now,
we also have irreducible implies prime.
Take a to be irreducible and assume that a [ bc. This means that bc = ad
for some d R. Using the property of unique factorization, we decompose d, b
and c into products of irreducible terms (resp. d
i
, b
i
, c
i
up to units u, v, w):
a ud
1
d
r
= vb
1
b
s
wc
1
. . . c
t
.
Since the factorization is unique, a must be associate to some either b
i
or c
i
,
implying that a divides b or c, which concludes the proof.
We now want to connect the property of unique factorization to ideals.
Denition 2.23. Let a
1
, a
2
, . . . be elements of an integral domain R. If the
sequence of principal ideals
(a
1
) (a
2
) (a
3
) . . .
stabilizes, i.e., we have
(a
n
) = (a
n+1
) = . . .
for some n, then we say that R satises the ascending chain condition on prin-
cipal ideals.
If the same condition holds but for general ideals, not necessarily principal,
we call R a Noetherian ring, in honor of the mathematician Emmy Noether.
Theorem 2.15. Let R be an integral domain.
1. If R is a UFD, then R satises the ascending chain condition on principal
ideals.
74 CHAPTER 2. RING THEORY
Figure 2.1: Amalie Emmy Noether (1882-1935)
2. If R satises the ascending chain condition on principal ideals, then every
non-zero element of R can be factored into irreducible (this says nothing
about the unicity of the factorization).
3. If R is such that every non-zero element of R can be factored into irre-
ducible, and in addition every irreducible element is prime, then R is a
UFD.
Thus R is a UFD if and only if it satises the ascending chain condition on
principal ideals and every irreducible element of R is prime.
Proof. 1. Recall that in a UFD, prime and irreducible are equivalent. Con-
sider an ascending chain of principal ideals
(a
1
) (a
2
) (a
3
) . . .
We have that a
i+1
[ a
i
for all i. Thus the prime factors of a
i+1
consist of
some (possibly all) prime factors of a
i
. Since a
1
has a unique factorization
into nitely many prime factors, the prime factors will end up being the
same, and the chain will stabilize.
2. Take 0 ,= a
1
R. If a
1
is irreducible, we are done. Let us thus assume
that a
1
is not irreducible, that is
a
1
= a
2
b
2
where a
2
and b
2
are not unit. Since a
2
[ a
1
, we have (a
1
) (a
2
), and
actually
(a
1
) (a
2
).
2.6. UNIQUE FACTORIZATION AND EUCLIDEAN DIVISION 75
Indeed, if (a
1
) = (a
2
), then a
2
would be a multiple of a
1
, namely a
2
= ca
1
and thus
a
1
= a
2
b
2
a
1
= ca
1
b
2
a
1
(1 cb
2
) = 0
implying that cb
2
= 1 and thus b
2
is a unit. This contradicts the as-
sumption that a
1
is not irreducible. This computation has shown us that
whenever we get a factor which is not irreducible, we can add a new princi-
pal ideal to the chain of ideals. Thus, if a
2
b
2
is a product of irreducible, we
are done. Otherwise, we have that say a
2
is not irreducible, and a
2
= a
3
b
3
,
yielding
(a
1
) (a
2
) (a
3
).
Since R satises the ascending chain condition on principal ideals, this
process cannot go on and must stop, showing that we have a factorization
into irreducible.
3. We now know that R allows a factorization into irreducible. We want to
prove that this factorization is unique, under the assumption that every
irreducible is prime. Suppose thus that
a = up
1
p
2
p
n
= vq
1
q
2
q
m
where u, v are units and p
i
, q
j
are irreducible. p
1
is an irreducible but also
a prime by assumption, thus it must divides one of the q
j
, say q
1
, and we
have q
1
= p
1
d. Since q
1
is irreducible, d must be a unit, and q
1
and p
1
are
associate. We can iterated the process to nd that q
i
and p
i
are associate
for all i.
We now introduce a notion stronger than being a unique factorization do-
main.
Denition 2.24. A principal ideal domain (PID) is an integral domain in which
every ideal is principal.
Theorem 2.16. A principal ideal domain R is a unique factorization domain.
Proof. What we will prove is that if R is a principal ideal domain, then
R satises the ascending chain condition on principle ideals.
every irreducible in R is also prime.
Having proved these two claims, we can conclude using the above theorem.
Let us rst prove that R satises the ascending chain condition on principle
ideals. Consider the following sequence of principal ideals
(a
1
) (a
2
) (a
3
) . . .
and let 1 =

i=1
(a
i
). Note that 1 is an ideal of R (be careful, a union of ideals
is not an ideal in general!). Indeed, we have that 1 is closed under addition:
76 CHAPTER 2. RING THEORY
take a, b 1, then there are ideals 1
j
and 1
k
in the chain with a 1
j
and
b 1
k
. If m max(j, k), then both a, b 1
m
and so do a + b. To check that
1 is closed under multiplication by an element of R, take again a 1. Then
a 1
j
for some j. If r R, then ra 1
j
implying that ra 1.
Now by assumption, 1 is a principal ideal, generated by, say b: 1 = (b).
Since b belongs to

i=1
(a
i
), it must belongs to some (a
n
). Thus 1 = (b) (a
n
).
For j n, we have
(a
j
) 1 (a
n
) (a
j
)
which proves that the chain of ideal stabilizes.
We are left to prove that every irreducible element is also prime. Let thus
a be an irreducible element. Consider the principal ideal (a) generated by a.
Note that (a) is a proper ideal: if (a) = R, then 1 (a) and thus a is a unit,
which is a contradiction.
We have that (a) is included in a maximal ideal 1 (this can be deduced from
either the ascending chain condition or from the theorem (Krulls theorem) that
proves that every ideal is contained in a maximal ideal). Since R is a principal
ideal domain, we have that 1 = (b). Thus
(a) (b) b [ a a = bd
where a is irreducible, b cannot be a unit (since 1 is by denition of maximal
ideal a proper ideal), and thus d has to be a unit of R. In other words, a and b
are associate. Thus
(a) = 1 = (b).
Since 1 is a maximal ideal, it is prime implying that a is prime, which concludes
the proof.
Determining whether a ring is a principal ideal domain is in general quite
a tough question. It is still an open conjecture (called Gausss conjecture) to
decide whether there are innitely many real quadratic elds which are principal
(we use the terminology principal for quadratic elds by abuse of notation, it
actually refers to their ring of integers, that is rings of the form either Z[

d] if
d 2 or 3 mod 4 or Z[
1+

d
2
] else).
One way mathematicians have found to approach this question is to actually
prove a stronger property, namely whether a ring R is Euclidean.
Denition 2.25. Let R be an integral domain. We say that R is a Euclidean
domain is there is a function from R0 to the non-negative integers such
that
a = bq +r, a, b R, b ,= 0, q, r R
where either r = 0 or (r) < (b).
When the division is performed with natural numbers, it is clear what it
means that r < b. When we work with polynomials instead, we can say that
deg r < deg b. The function generalizes these notions.
2.6. UNIQUE FACTORIZATION AND EUCLIDEAN DIVISION 77
Theorem 2.17. If R is a Euclidean domain, then R is a principal ideal domain.
Proof. Let 1 be an ideal of R. If 1 = 0, it is principal and we are done. Let
us thus take 1 , = 0. Consider the set
(b), b 1, b ,= 0.
It is included in the non-negative integers by denition of , thus it contains a
smallest element, say n. Let 0 ,= b 1 such that (b) = n.
We will now prove that 1 = (b). Indeed, take a 1, and compute
a = bq +r
where r = 0 or (r) < (b). This yields
r = a bq 1
and (r) < (b) cannot possibly happen by minimality of n, forcing r to be
zero. This concludes the proof.
Example 2.7. Consider the ring
Z[

d] = a +b

d, a, b Z
with
(a +b

d) = [a
2
bd
2
[.
We will show that we have a Euclidean domain for d = 2, 1, 2, 3.
Note that Z[

d] is an integral domain. Take , ,= 0 in Z[

d]. Now we
perform the division of by to get something of the form x +

dy. Since
Z[

d] is not a eld, there is no reason for this division to give a result in Z[

d]
(that is, x, y Z), however, we can compute it in Q(

d), and get a result with

x, y rational. Since x, y have no reason to be integers, let us approximate them
by integers x
0
, y
0
, namely take x
0
, y
0
such that
[x x
0
[ < 1/2, [y y
0
[ < 1/2.
Let
q = x
0
+y
0

d, r = ((x x
0
) + (y y
0
)

d)
then
q +r = (x
0
+y
0

d) +((x x
0
) + (y y
0
)

d)
= (x +y

d) = .
We are left to show that () < (). We have
(r) = ()((x x
0
) + (y y
0
)

d)
= ()[(x x
0
)
2
d(y y
0
)
2
]
< ()
_
1
4
+[d[
1
4
_
showing that Z[

d] is indeed a Euclidean domain for d = 2, 1, 2, 3.

78 CHAPTER 2. RING THEORY
Below is a summary of the ring hierarchy (recall that PID and UFD stand
respectively for principal ideal domain and unique factorization domain):
integral domains UFD PID Euclidean domains
Note that though the Euclidean division may sound like an elementary con-
cept, as soon as the ring we consider is fancier than Z, it becomes quickly
a dicult problem. We can see that from the fact that being Euclidean is
stronger than being a principal ideal domain. All the inclusions are strict, since
one may check that Z[

3] is an integral domain but is not a UFD, Z[X] is a

UFD which is not PID, while Z[(1+i

19)/2] is a PID which is not a Euclidean

domain.
2.7 Irreducible polynomials
Recall the denition of irreducible that we have seen: a non-zero element a
which is not a unit is said to be irreducible if a = bc implies that either b or c
is a unit. Let us focus on the case where the ring is a ring of polynomials R[X]
and R is an integral domain.
Denition 2.26. If R is an integral domain, then an irreducible element of
R[X] is called an irreducible polynomial.
In the case of a eld F, then units of F[X] are non-zero elements of F.
Then we get the more familiar denition that an irreducible element of F[X] is
a polynomial of degree at least 1, that cannot be factored into two polynomials
of lower degree.
Let us now consider the more general case where R is an integral domain
(thus not necessarily a eld, it may not even be a unique factorization domain).
To study when polynomials over an integral domain R are irreducible, it is
often more convenient to place oneselves in a suitable eld that contains R,
since division in R can be problematic. To do so, we will now introduce the
eld of fractions, also called quotient eld, of R. Since there is not much more
diculty in treating the general case, that is, when R is a commutative ring,
we present this construction.
Let S be a subset of R which is closed under multiplication, contains 1 and
does not contain 0. This denition includes the set of all non-zero elements of
an integral domain, or the set of all non-zero elements of a commutative ring
that are not zero divisors. We dene the following equivalence relation on RS:
(a, b) (c, d) s(ad bc) = 0 for some s S.
It is clearly reexive and symmetric. Let us check the transitivity. Suppose that
(a, b) (c, d) and (c, d) (e, f). Then
s(ad bc) = 0 and t(cf de) = 0
2.7. IRREDUCIBLE POLYNOMIALS 79
for some s, t S. We can now multiply the rst equation by tf, the second by
sb and add them
stf(ad bc) +tsb(cf de) = 0
to get
sdt(fa be) = 0
which proves the transitivity.
What we are trying to do here is to mimic the way we deal with Z. If we take
non-zero a, b, c, d Z, we can write down a/b = c/d, or equivalently ad = bc,
which is also what (a, b) (c, d) satises by denition if we take R to be an
integral domain. In a sense, (a, b) is some approximation of a/b.
Formally, if a R and b S, we dene the fraction a/b to be the equivalence
class of the pair (a, b). The set of all equivalence classes is denoted by S
1
R.
To make it into a ring, we dene the following laws in a natural way:
addition:
a
b
+
c
d
=
ad +bc
bd
.
multiplication:
a
b
c
d
=
ac
bd
.
additive identity:
0
1
=
0
s
, s S.
additive inverse:

a
b
=
a
b
.
multiplicative identity:
1
1
=
s
s
, s S.
To prove that we really obtain a ring, we need to check that all these laws
are well-dened.
Theorem 2.18. With the above denitions, the set of equivalence classes S
1
R
is a commutative ring.
1. If R is an integral domain, so is S
1
R.
2. If R is an integral domain, and S = R0, then S
1
R is a eld.
Proof. Addition is well-dened. If a
1
/b
1
= c
1
/d
1
and a
2
/b
2
= c
2
/d
2
, then
for some s, t S, we have
s(a
1
d
1
b
1
c
1
) = 0 and t(a
2
d
2
b
2
c
2
) = 0.
We can now multiply the rst equation by tb
2
d
2
and the second by sb
1
d
1
to get
tb
2
d
2
s(a
1
d
1
b
1
c
1
) = 0 and sb
1
d
1
t(a
2
d
2
b
2
c
2
) = 0,
80 CHAPTER 2. RING THEORY
and adding them yields
st[d
2
d
1
(b
2
a
1
+b
1
a
2
) b
2
b
1
(d
2
c
1
+d
1
c
2
)] = 0
that is
b
2
a
1
+b
1
a
2
b
2
b
1
=
d
2
c
1
+d
1
c
2
d
2
d
1
,
which can be rewritten as
a
1
b
1
+
a
2
b
2
=
c
1
d
1
+
c
2
d
2
and we conclude that addition does not depend on the choice of a representative
in an equivalence class.
Multiplication is well-dened. We start as before. If a
1
/b
1
= c
1
/d
1
and
a
2
/b
2
= c
2
/d
2
, then for some s, t S, we have
s(a
1
d
1
b
1
c
1
) = 0 and t(a
2
d
2
b
2
c
2
) = 0.
Now we multiply instead the rst equation by ta
2
d
2
, the second by sc
1
b
1
and
we add them:
st[a
2
d
2
a
1
d
1
c
1
b
1
b
2
c
2
] = 0.
This implies, as desired, that
a
1
a
2
b
1
b
2
=
c
1
c
2
d
1
d
2
.
To be complete, one should check that the properties of a ring are fullled, but
this follows from the fact that addition and multiplication are carried the usual
way.
1. We want to prove that S
1
R is an integral domain. We assume that R
is an integral domain, and we need to check the denition of an integral
domain. Namely, suppose that (a/b)(c/d) = 0 in S
1
R, that is
a
b
c
d
=
0
1
.
This means that (ac, bd) (0, 1) and acs = 0 for some s S. Now acs = 0
is an equation in R, which is an integral domain, and s ,= 0, thus ac = 0,
so either a or c is 0, and consequently either a/b or c/d is zero.
2. To conclude, we want to prove that S
1
R is a eld, assuming that R is
an integral domain, and S = R0. We consider a/b a non-zero element
of S
1
R, for which we need to nd an inverse. Note that a and b are
non-zero, thus they are both in S meaning that both a/b and b/a are in
S
1
R and b/a is the multiplicative inverse of a/b.
2.7. IRREDUCIBLE POLYNOMIALS 81
Denition 2.27. Let R be a commutative ring. Based on the above, the set
of equivalence classes S
1
R is a commutative ring, called the ring of fractions
of R by S. If R is an integral domain, and S = R0, then S
1
R is called the
eld of fractions or quotient eld of R.
Now that we have dened a suitable eld, we are left to prove that we can
embed an integral domain R in its quotient eld.
Proposition 2.19. A commutative ring R can be embedded in its ring of frac-
tions S
1
R, where S is the set of all its non-divisors of zero. In particular, an
integral domain can be embedded in its quotient eld, which is furthermore the
smallest eld containing R.
Proof. Consider the following map:
f : R S
1
R, a f(a) = a/1.
It is not hard to check that f is a ring homomorphism. If S has no zero divisor,
we have that the kernel of f is given by the set of a such that f(a) = a/1 = 0/1,
that is the set of a such that sa = 0 for some s. Since s is not a zero divisor,
we have a = 0 and f is a monomorphism.
Let us get back to the irreducible polynomials, and consider now the case
where D is a unique factorization domain. It is not necessarily a eld, but we
now know how to embed it in a suitable eld, namely its eld of fractions, or
quotient eld. Take the polynomial f(X) = a + abX, a ,= 0 not a unit. Since
we can factor it as
f(X) = a(1 +bX)
where a is not a unit by assumption, this polynomial is not irreducible. But we
do not really have a factorization into two polynomials of lower degree. What
happens here is that the constant polynomials are not necessarily units, unlike in
the case of elds. To distinguish this case, we introduce the notion of primitive
polynomial.
Denition 2.28. Let D be a unique factorization domain and let f D[X].
We call the greatest common divisor of all the coecients of f the content of
f, denoted by c(f). A polynomial whose content is a unit is called a primitive
polynomial.
We can now rule out the above example, and we will prove later that this
allows us to say that a primitive polynomial is irreducible if and only if it
cannot be factored into two polynomials of lower degree. Be careful however
that primitive polynomial has a dierent meaning if it is dened over a eld.
The next goal is to prove Gauss lemma, which in particular implies that the
product of two primitive polynomials is a primitive polynomial.
We start with a lemma.
Lemma 2.20. Let D be a unique factorization domain, and consider f ,=
0, g, h D[X] such that pf(X) = g(X)h(X) with p a prime. Then either p
divides all the coecients of g or p divides all the coecients of h.
82 CHAPTER 2. RING THEORY
Figure 2.2: Carl Friedrich Gauss (1777-1855)
Before starting the proof, let us notice that this lemma is somehow a gener-
alization of the notion of prime. Instead of saying that p[ab implies p[a or p[b, we
have p[g(X)h(X) implies that p[g(X) or p[h(X) (dividing the whole polynomial
means dividing all of its coecients).
Proof. Denote
g(X) = g
0
+g
1
X +. . . +g
s
X
s
, h(X) = h
0
+h
1
X +. . . +h
t
X
t
.
Suppose by contradiction that p does not divide all coecients of g and does
not divide all coecients of h either. Then let g
u
and h
v
be the coecients of
minimum index not divisible by p. Then the coecient of X
u+v
in g(X)h(X)
is
g
0
h
u+v
+g
1
h
u+v1
+. . . +g
u
h
v
+. . . +g
u+v1
h
1
+g
u+v
h
0
.
By denition of u and v, p divides every term but g
u
h
v
, thus p cannot possibly
divide the entire expression, and thus there exists a coecient of g(X)h(X) not
divisible by p. This contradicts the fact that p[g(X)h(X).
Proposition 2.21. (Gauss Lemma). Let f, g be non-constant polynomials
in D[X] where D is a unique factorization domain. The content of a product of
polynomials is the product of the contents, namely
c(fg) = c(f)c(g),
up to associates. In particular, the product of two primitive polynomials is
primitive.
Proof. Let us start by noticing that by denition of content, we can rewrite
f(X) = c(f)f

(X), g(X) = c(g)f

(X),
2.7. IRREDUCIBLE POLYNOMIALS 83
where f

, g

D[X] are primitive. Clearly

fg = c(f)c(g)f

.
Since c(f)c(g) divides fg, it divides every coecient of fg and thus their
greatest common divisor:
c(f)c(g) [ c(gf).
We now prove the converse, namely that c(gf)[ [c(g)c(g). To do that, we
consider each prime p appearing in the factorization of c(gf) and argue that
p [ c(f)c(g). Let thus p be a prime factor of c(gf). Since fg = c(fg)(fg)

, we
have that c(fg) divides fg, that is
p [ c(f)c(g)f

= (c(f)f

)(c(g)g

).
By the above lemma, either p [ c(f)f

or p [ c(g)g

, say p [ c(f)f

, meaning
that either p [ c(f) or p [ f

. Since f

is primitive, p cannot possibly divide f

,
and thus
p [ c(f) p [ c(f)c(g).
If p appears with multiplicity, we iterate the reasoning with the same p.
We are now ready to connect irreducibility over a unique factorization do-
main and irreducibility over the corresponding quotient eld or eld of fractions.
Proposition 2.22. Let D be a unique factorization domain with quotient eld
F. If f is a non-constant polynomial in D[X], then f is irreducible over D if
and only if f is primitive and irreducible over F.
Proof. First assume that f is irreducible over D.
f is primitive. Indeed, if f were not primitive, then we could write
f = c(f)f

,
where c(f) denotes the content of f and f

is primitive. Since we assume f is

not primitive, its content cannot be a unit, which contradicts the irreducibility
of f over D, and we conclude that f is primitive.
f is irreducible over F. Again assume by contradiction that f is not
irreducible over F. Now F is a eld, thus reducible means f can be factored
into a product of two non-constant polynomials in F[X] of smaller degree:
f(X) = g(X)h(X), deg g < deg f, deg h < deg f.
Since g, h are in F[X], and F is the eld of fractions of D, we can write
g(X) =
a
b
g

(X), h(X) =
c
d
h

(X), a, b, c, d D
and g

, h

primitive. Thus
f(X) =
ac
bd
g

(X)h

(X)
84 CHAPTER 2. RING THEORY
Figure 2.3: Ferdinand Eisenstein (1823-1852)
where g

is a primitive polynomial by Gauss Lemma. Since we have already

proved that f

is primitive, it must be that bd = ac. But this would mean that

f(X) = g

(X)h

(X)
which contradicts the fact that f(X) is irreducible over D[X] and we conclude
that f is also irreducible over F[X].
We are left to prove the converse. Let then f be a primitive and irreducible
polynomial over F. We do it by contraction, and assume that the primitive
polynomial f is not irreducible over D:
f(X) = g(X)h(X).
Since f is primitive, deg g and deg h are at least 1. But then neither g not h
can be a unit in F[X] (these are units in F) and thus
f = gh
contradicts the irreducibility of f over F.
In other words, we have proven that f irreducible over D is equivalent to f
primitive and cannot be factored into two polynomials of lower degree in F[X].
To conclude, we present a practical criterion to decide whether a polynomial
in D[X] is irreducible over F.
2.7. IRREDUCIBLE POLYNOMIALS 85
Proposition 2.23. (Eisensteins criterion). Let D be a unique factorization
domain, with quotient eld F and let
f(X) = a
n
X
n
+. . . +a
1
X +a
0
be a polynomial in D[X] with n 1 and a
n
,= 0.
If p is a prime in D and p divides a
i
, 0 i n but p does not divide a
n
nor does p
2
divide a
0
, then f is irreducible over F.
Proof. We rst divide f by its content, to get a primitive polynomial. By
the above proposition, it is enough to prove that this primitive polynomial is
irreducible over D.
Let thus f be a primitive polynomial and assume by contradiction it is
reducible, that is
f(X) = g(X)h(X)
with
g(X) = g
0
+. . . +g
r
X
r
, h(X) = h
0
+. . . +h
s
X
s
.
Notice that r cannot be zero, for if r = 0, then g
0
= g would divide f and
thus all a
i
implying that g
0
divides the content of f and is thus a unit. But this
would contradict the fact that f is reducible. We may from now on assume that
r 1, s 1.
Now by hypothesis, p [ a
0
= g
0
h
0
but p
2
does not divide a
0
, meaning that p
cannot divide both g
0
and h
0
. Let us say that
p [ g
0
and p does not divide h
0
(and vice-versa).
By looking at the dominant coecient a
n
= g
r
h
s
, we deduce from the as-
sumption that p does not divide a
n
that p cannot possibly divide g
r
. Let i be
the smallest integer such that p does not divide g
i
. Then
1 i r < n = r +s.
Let us look at the ith coecient
a
i
= g
0
h
i
+g
1
h
i1
+. . . +g
i
h
0
and by choice of i, p must divide g
0
, . . . , g
i1
. Since p divides a
i
by assumption,
it thus must divide the last term g
i
h
0
, and either p [g
i
or p [ h
0
by denition of
prime. Both are impossible: we have chosen p dividing neither h
0
nor g
i
. This
concludes the proof.
86 CHAPTER 2. RING THEORY
The main denitions and results of this chapter are
(2.1-2.2). Denitions of: ring, zero divisor, unit,
integral domain, division ring, subring, characteristic,
ring homomorphism, ideal, quotient ring. Factor and
1st Isomorphism Theorem for rings.
(2.3-2.4). Operations on ideals, Chinese Remainder
Theorem, Correspondence Theorem for rings. Deni-
tions of: principal ideal, maximal ideal, prime ideal,
the characterization of the two latter in the commu-
tative case.
(2.5). Polynomial Euclidean division, number of
roots of a polynomial.
(2.6). Denitions of: associate, prime, irreducible,
unique factorization domain, ascending chain condi-
tion, principal ideal domain, Euclidean domain. Con-
nections between prime and irreducible. Hierarchy
among UFD, PID and Euclidean domains.
(2.7). Construction of ring of fractions. Denitions
of: content of a polynomial, primitive polynomial.
Gauss Lemma, Eisensteins criterion.