Unique Factorization

Wae
Mathcamp 2010
Throughout these notes, all rings will be assumed to be commutative.
1 Factorization in domains: denitions and examples
In this class, we will study the phenomenon of unique factorization in rings, as a generalization of unique
factorization of integers into primes. First, well dene what unique factorization means in general.
Denition 1.1. A (commutative) ring R is a domain (or integral domain) if whenever ab = 0 for a, b R,
then either a = 0 or b = 0. An element u R is a unit if it has an inverse, i.e. if there exists u
1
R such
that uu
1
= 1. If every nonzero element of R is a unit, we say R is a eld.
Domains will be the basic setting in which it will make sense to ask questions about unique factorization.
Units are important because they are the ambiguity in unique factorizations. For example, in Z, the only
units are 1 and 1. The fact that 1 is a unit means that factorization of integers into primes isnt really
unique. For example, we can factorize 6 = 2 3 or 6 = (2) (3). However, we still consider these two
factorizations to be the same since 2 and 2 (respectively 3 and 3) only dier by multiplication by the
unit 1.
Denition 1.2. Let R be a domain, and a, b R. We write a|b (a divides b, or a is a factor of b) if there
exists c such that b = ac. We say a nonzero element p R is irreducible if it is not a unit and if whenever
a|p, either a is a unit or a = up for some unit u. Alternatively, whenever we can write p = ab, one of a and
b must be a unit.
Note that any factor of a unit is again a unit. Indeed, if u = ab, then abu
1
= 1 so bu
1
is an inverse for
a.
We nally dene unique factorization itself.
Denition 1.3. A domain R has unique factorization (or is a unique factorization domain UFD) if every
nonzero a R can be written uniquely as a product
a = u

p
di
i
where u is a unit and the p
i
are irreducible. Here uniquely means up to reordering the factors and
multiplying the factors by units.
Whew! Thats a lot of denitions. Lets look at some examples.
Example 1.4. The integers Z are a domain but not a eld. The only units in Z are 1 and 1. An integer
is irreducible i it is either a prime number or negative a prime number. By the fundamental theorem of
arithmetic (which we will prove later), Z is a UFD.
Example 1.5. The rationals Q, the reals R, and the complex numbers C are all elds. Any eld has no
irreducibles, and is automatically a UFD
1
Example 1.6. Any subring of a domain is a domain. In particular, any collection of complex numbers
closed under addition, subtraction, and multiplication is a domain. For example, for any integer n Z,
there is a domain
Z[

n] = {a + b

n : a, b Z}
since it can be checked that such numbers are closed under addition, subtraction, and multiplication.
Example 1.7. Lets look more closely at a special case of the example above, namely Z[

1] = Z[i], also
known as the Gaussian integers. The Gaussian integers are a lot like the integers, but some things work a
little dierently. First of all, besides 1 and 1, i and i are also units since i (i) = 1. Also, for instance, 2
is irreducible as an integer, but not as a Gaussian integer, since we can factor it as 2 = (1 + i)(1 i). More
generally, if p is a prime which can be written as a sum of two squares p = a
2
+ b
2
, then p is not irreducible
as a Gaussian integer because it factors p = (a + bi)(a bi).
To say much more about factorization in Z[i], we need to use the norm N(a + bi) = |a + bi|
2
= a
2
+ b
2
,
a map from Z[i] to the nonnegative integers. The key property of the norm is that N(xy) = N(x)N(y). In
particular, if u is a unit, then
1 = N(1) = N(uu
1
) = N(u)N(u
1
),
which implies N(u) must be 1. It follows that 1 and i are the only units in Z[i].
Similarly, let p be a prime, and suppose that in Z[i], we can factor p = xy in a nontrivial way. Then
N(p) = p
2
= N(x)N(y). Since x and y are not units, we must have N(x) = N(y) = p, so p must be a sum
of two squares (say if x = a + bi, then p = a
2
+ b
2
). Furthermore, the same argument shows that x and
y themselves must be irreducible. On the other hand, if p cannot be written as a sum of two squares, it is
irreducible in Z[i].
We have now gured out how to factor any ordinary integer n into irreducible Gaussian integers. Namely,
for each ordinary prime factor p, if p is not a sum of two squares, it is irreducible, and if p = a
2
+ b
2
, then
we can replace p in the factorization of n with (a + bi)(a bi), and each of these is irreducible.
It turns out that Z[i] is a UFD; we will prove this later. As a consequence, we can show that in fact the
above tells us everything about factorization in Z[i]: the irreducibles of Z[i] are exactly the primes p which
are not sums of two squares and a bi such that a
2
+ b
2
is prime. To show this, let x = a + bi Z[i]. Then
N(x) = a
2
+ b
2
= x x. We know that all the irreducible factors of N(x) are of the form mentioned above,
since N(x) is an ordinary integer. By unique factorization, every irreducible factor of x is also one of N(x).
Thus every irreducible factor of x must be of the form above, and so every irreducible Gaussian integer is of
the form above.
Example 1.8. Almost everything in the discussion above also applies to the example Z[

5] (or of Z[

n]
for any n). The only dierence is that the norm is given by N(a + b

5) = a
2
+ 5b
2
, so everywhere we
said sum of two squares we should now say of the form a
2
+ 5b
2
. However, there is one big dierence
between Z[

5] and Z[i]: Z[

5] is not a UFD!
To see this, lets factor 9 in Z[

5]. Well, rst of all 9 = 3 3, and 3 is irreducible since it is not of

the form a
2
+ 5b
2
. On the other hand, 9 = 2
2
+ 5 1
2
, so 9 = (2 +

5)(2

5). Note that 2

5
is irreducible for the same reason 3 is: it has norm 9, and no factor can have norm 3 since 3 is not of the
form a
2
+ 5b
2
. Clearly 3 does not dier from 2

5 by a unit, so we conclude that 9 has two dierent

factorizations in this ring!
1.1 Exercises
Exercise 1.1. Factor 14 + 6i into irreducible Gaussian integers. (Hint: Use the norm.)
Exercise 1.2. Let R be a UFD and let p R be irreducible. Show that if p|ab for some a, b R, then
either p|a or p|b. (Tomorrow, we will show that this property is basically equivalent to R being a UFD).
Exercise 1.3. Show that Z[2i] = {a + bi : a, b Z and b is even} is not a UFD.
Exercise 1.4. Show that Z[

6] is not a UFD. (Hint: Find an example similar to the example for Z[

5].)
2
Exercise 1.5. Let k be a eld and let R = k[x, y, z, w]/(xy zw), the quotient of the polynomial ring
k[x, y, z, w] in four variables by the ideal generated by xy zw. Assuming that R is a domain (which we
will prove later), show that it does not have unique factorization. (Hint: xy = zw looks like two dierent
factorizations of the same element. Can you show that it really is?)
Exercise 1.6. Let R be a domain and a, b R be nonzero. Then a greatest common divisor (or gcd) of a
and b is a d R such that d|a, d|b, and whenever c|a and c|b, c|d.
(a): Show that the gcd of two elements, if it exists, is unique up to multiplication by a unit.
(b): Suppose R is a UFD. Show that any two elements have a gcd.
3
2 Primes and PIDs
We start by looking at a few more examples of the sorts of domains in which we might ask about unique
factorization.
Example 2.1. Let R be a ring. Then we can form the ring R[x] of polynomials (in one variable) with
coecients in R. A polynomial is a formal expression a
n
x
n
+ a
n1
x
n1
+ . . . + a
1
x + a
0
where a
i
R,
and polynomials are added and multiplied according to the familiar rules from elementary algebra. If
f(x) =

a
k
x
k
, the largest n such that a
n
= 0 is the degree deg(f) of f, and we call a
n
the leading
coecient of f.
1
If R is a domain, then deg(fg) = deg(f) + deg(g) since the leading coecients of f and
g multiply to give a nonzero leading coecient of fg. It follows that if R is a domain, so is R[x]. It also
follows that the units of R[x] are just the units of R, since any unit must have degree 0.
Example 2.2. Lets specialize the example above to the case that k is a eld; in this case we can assume all
polynomials are monic (leading coecient 1) by multiplying by a unit. The simplest irreducible polynomials
are linear polynomials x a; these are necessarily irreducible for reasons of degree. It is a consequence of
polynomial division that (x a)|f(x) i f(a) = 0; we will say more about this later. This means that if a
polynomial of degree > 1 has a root, it is reducible. In particular, the fundamental theorem of algebra says
that if k = C, every (nonconstant) polynomial has a root, so the only irreducible polynomials are the linear
polynomials. On the other hand, there are other irreducible polynomials over other elds. For example,
if k = R or k = Q, x
2
+ 1 is irreducible: a factor of it would have to be linear and thus give a root, but
x
2
+ 1 has no root. A similar argument shows that if k = Q, then x
3
2 is irreducible. Note, though, that
a polynomial can have no roots but still be reducible. For example, x
4
4 = (x
2
2)(x
2
+ 2) has no roots
for k = Q.
We will say more about factorization of polynomials later. It turns out that for any eld k, k[x] is a UFD.
In fact, more generally, if R is a UFD, so is R[x]. It follows by induction that a polynomial ring k[x, y, z, . . . ]
in any (nite) number of variables is a UFD.
We now look at how you might prove that a domain has unique factorization. The one example where
we already know this is Z, so lets analyze the proof in that case.
Theorem 2.3 (Fundamental theorem of arithmetic). Z is a UFD.
Proof. Let n Z be nonzero; by multiplying n by a unit we may assume n > 1. We rst show the existence
of a factorization of n by induction on n. The base case n = 1 is trivial; we simply have no factors in the
product.
Now suppose that every positive m < n has a factorization. If n is prime, it clearly has a factorization
with itself as the only factor. Otherwise, we can factor n = ab where neither a nor b is a unit, and we may
also assume that a, b > 0. It follows that a, b < n, so by induction a and b have factorizations. By combining
the factorizations of a and b together, we obtain a factorization of n.
Now we show the factorization is unique. For this, we need a lemma we will come back to later.
Lemma 2.4. Let p Z be prime and suppose p|ab. Then p|a or p|b.
Assuming the Lemma, the proof of uniqueness is not hard. Suppose n = u

p
di
i
and n = v

q
ej
j
are
two factorizations. We can one-by-one match up the p
i
with the q
j
: since p
1
|n = v

q
ej
j
, by Lemma 2.4 p
1
must divide either v or one of the q
j
. Since a factor of a unit is a unit, p
1
cannot divide v, so it must divide
some q
j
. Since q
j
is irreducible and p
1
is not a unit, p
1
must be the same as q
j
, up to a unit.
Having matched up p
1
and q
j
, we can remove them from the respective factorizations and then repeat,
pairing up another two factors. Repeating this, we see that every factor in the rst factorization must also
appear in the second (up to a unit), and vice versa. Thus the two factorizations are equivalent.
Note that once we had Lemma 2.4, the proof of uniqueness used nothing special about Z: it would hold
in any domain that satised Lemma 2.4. Lets turn Lemma 2.4 into a denition.
1
We formally write deg(0) = .
4
Denition 2.5. A nonzero nonunit element p of a domain is prime if whenever p|ab, p|a or p|b.
Note that primes are automatically irreducible: indeed, if p factors as ab, then p|ab, so (say) p|a and
a = pc Then p = ab = pbc, so bc = 1, so b is a unit. Furthermore, in a UFD, irreducibles are prime: if
p|ab, a factorization of ab is given by combining factorizations of a and b, so p must divide one of them by
uniqueness. In fact, this is actually equivalent to unique factorization!
Theorem 2.6. Let R be a domain such that every nonzero a R has a factorization into irreducible
elements. Then if every irreducible element of R is prime, R is a UFD.
Proof. The proof is exactly the same as the proof above in the case Z = R.
Thus the hard work is really in showing that irreducibles are prime, i.e. Lemma 2.4.
2
We now give a
proof of this in the case of R = Z that will generalize to many other cases. We rst need a denition.
Denition 2.7. Recall that an ideal in a ring R is a subset I R closed under addition such that for any
a R and i I, ai I. For any a R, we have an ideal (a) = {ar : r R} = {b : a|b}; we call such ideals
principal. If every ideal in a domain R is principal, we say R is a principal ideal domain (PID).
Here is one concrete consequence of being a PID. Let a, b R. Then the ideal (a, b) = {ar +bs : r, s R}
must be principal; say (a, b) = (c). Then a, b (c), so c|a and c|b. Furthermore, we can write c = ar + bs
for some r and s, which implies that every other common factor of a and b also divides c. That is, c is a
greatest common divisor (gcd) of a and b, and is a combination ar + bs of a and b.
Theorem 2.8. In a PID, irreducibles are prime.
Proof. Let R be a PID and p R be irreducible. Suppose p|ab and p |b; we will show p|a. Consider the
ideal (p, b) = (c), where c is a gcd of p and b. Then c|p, so since p is irreducible, either c is a unit or c is the
same as p, up to a unit. If c is the same as p, then (p, b) = (c) = (p) so b (p) and p|b, contradicting our
assumption. Thus c is a unit, and so we may assume c = 1.
We can thus write 1 = rp + sb for some r and s. Multiplying this equation by a, we get a = rpa + sba.
But p|ab, so p divides the right-hand side, so p|a, as desired.
To complete the proof of the fundamental theorem of arithmetic, we thus must only prove:
Theorem 2.9. Z is a PID.
Proof. Let I Z be an ideal. If I = {0}, then I = (0) is principal. Otherwise, I contains some nonzero
n, and by multiplying by 1, I contains some n > 0. Let n be the smallest positive element of I. Then
I (n). For any m I, by division with remainder, we can write m = qn+r where 0 r < n. Since m I
and qn I, it follows that r I. But r < n, so by minimality of n we must have r = 0. Thus m = qn is
divisible by n, so I = (n).
We now have a complete proof that Z is a UFD. More importantly, however, we have a machine set up to
show that other domains are also UFDs: we just have to show they are PIDs in which factorizations always
exist. Tomorrow, we will use this to prove that other rings are UFDs.
2.1 Exercises
Exercise 2.1. Let R be a domain and a R be irreducible. Show that a is also irreducible as an element
of the polynomial ring R[x].
Exercise 2.2. Let k[x
2
, x
3
] k[x] be the ring of polynomials with coecients in a eld k with no linear
term (which is the subring generated by k, x
2
, and x
3
). Show that k[x
2
, x
3
] is not a UFD. (Hint: Show that
x
2
and x
3
are irreducible.)
2
We also need to show that factorizations exist at all, but this is usually easy by some sort of induction argument, as we
gave for Z.
5
Exercise 2.3. Let k be a eld. Show that in k[x], every nonzero element can be factored (maybe not
uniquely) as u

p
di
i
where u is a unit and the p
i
are irreducible. (Hint: Use induction on the degree of the
polynomial.)
Exercise 2.4. Let k be a eld. Can you prove that k[x] is a PID by imitating the proof for Z? (We will
talk about this in class tomorrow.)
Exercise 2.5. Let k be a eld. Show that in k[x
1
, . . . , x
n
], every nonzero element can be factored (maybe
not uniquely) as u

p
di
i
where u is a unit and the p
i
are irreducible. (Hint: Dene a notion of degree for a
polynomial in multiple variables.)
Exercise 2.6. An ideal I R in a ring R is called prime if the quotient ring R/I is a domain.
(a): Show that I is prime i whenever ab I, either a I or b I.
(b): Show that an element a R is prime i the ideal (a) is a prime ideal.
Exercise 2.7. An ideal I R in a ring is called maximal if the quotient ring R/I is a eld.
(a): Show that I is maximal i any ideal J I containing I is equal to either I or R. (Hint: Show that a
ring is a eld i it has no ideals other than {0} and the whole ring.)
(b): Show that maximal ideals are prime.
(c): Show that in a PID, prime ideals (other than {0}) are maximal. (Hint: Note that (a) (b) i b|a.)
Exercise 2.8 (Chinese remainder theorem). Let R be a PID, and let a, b R with gcd(a, b) = 1. Dene a
map f : R/(ab) R/(a) R/(b) by f(x + (ab)) = (x + (a), x + (b)). Show that f is a ring isomorphism.
3
(For example, for R = Z, this says that if m and n are relatively prime, then Z/mn

= Z/mZ/n.)
3
If R and S is a ring, then R S is also a ring by adding and multiplying separately on each coordinate. In particular,
R/(a) R/(b) is a ring.
6
3 Euclidean domains
The argument we made yesterday for Z readily generalizes to some other domains.
Theorem 3.1. Let k be a eld. Then k[x] is a PID.
Proof. We do the same thing as we did for Z, except instead of taking a minimal positive element, we take
an element of minimal degree. More precisely, let I k[x] be an ideal, which we can assume to contain
some nonzero element. Let f(x) I be nonzero of minimal degree and let g(x) I; we will show that f|g
and hence I = (f). By polynomial long division, we can divide g = qf +r where the remainder r has degree
strictly less than the degree of f. Since g I and qf I, we also have r I. But by minimality of the
degree of f, this can only happen if r = 0. Thus g = qf so I = (f).
Corollary 3.2. Let k be a eld. Then k[x] is a UFD.
Proof. We only need the existence of factorizations. This can be proven by induction on degree, or using
the fact that we will prove later that in fact PIDs always have factorizations and hence are automatically
UFDs.
We now make a denition which generalizes the arguments above for Z and k[x].
Denition 3.3. A Euclidean domain is a domain R equipped with a function d : R {0} N such that:
For any nonzero a, b R, d(ab) max(d(a), d(b)), and
For any nonzero a, b R, we can write a = bq + r where either r = 0 or d(r) < d(b).
Example 3.4. Z is a Euclidean domain with d(n) = |n|.
Example 3.5. k[x] is a Euclidean domain with d(f) = deg(f).
Theorem 3.6. Euclidean domains are PIDs (and hence UFDs).
Proof. Let R be a Euclidean domain. The proof that R is a PID is exactly the same as the proof for R = Z
or R = k[x]: given an ideal I, take a I such that d(a) is minimized, and then a will generate I.
We can nally give an application of all this machinery: a quick proof that Z[i] is a UFD.
Theorem 3.7. Z[i] is a Euclidean domain with d(a + bi) = N(a + bi) = a
2
+ b
2
.
Proof. The rst condition on d is easy to check. For the second condition, let x, y Z[i]; we wish to divide
y by x with a remainder which has smaller norm than x. That is, we want to nd a multiple of x which
is closer to y than x is to 0. For this, we use the geometry of Gaussian integers in the complex plane: the
multiples of x form a square lattice, and elementary geometry shows that any point in the plane is closer to
some vertex of the lattice than the side length of the squares.
Corollary 3.8. Z[i] is a UFD.
This proof also can be used to see what goes wrong for Z[

5]. In that case, instead of having a square

lattice we have a rectangular lattice, and the geometric fact we used is no longer true.
In the previous section, we made a geometric argument that the norm made Z[i] a Euclidean domain, but
this argument failed for Z[

5]. Lets recall that argument: the point is that to get division with a smaller
remainder, you need that in a square with side length 1, every point is distance < 1 from some corner. In
the case of Z[

5], instead of a square we had a rectangle, with one side having length 1 and the other side
having length

5. More generally, if we had done Z[

d] for any d > 0, the other side would have length

d. A little geometry shows that the maximum distance a point can be from a corner is

d + 1/2. Thus
this argument still works for d = 2 (Exercise 3.2), breaks down for d > 3, and just barely fails for d = 3.
We now turn to the general question of the existence of factorizations in PIDs.
7
Lemma 3.9. Let R be a PID and a R be a non-unit. Then a has an irreducible factor.
Proof. Suppose a = a
0
has no irreducible factor. Then we can pick a nontrivial factor a
1
|a
0
, and a
1
cannot
have an irreducible factor either. Thus we can pick an innite sequence . . . a
2
|a
1
|a
0
, where each a
n
is a
nontrivial factor of a
n1
. This means exactly that (a
0
) (a
1
) (a
2
) . . . with each inclusion being strict.
Now consider the ideal I = (a
0
, a
1
, . . . ) generated by all the a
i
. Since R is a PID, I is generated by some
single element c. This c can be written as a nite linear combination of the a
i
, say c =

n
b
i
a
i
. But then
c (a
n
) since a
n
|a
i
for every i n. But then we have (a
n
) I = (c) (a
n
), which is a contradiction.
Hence no such a = a
0
could have existed, so any element has an irreducible factor.
Theorem 3.10. Let R be a PID. Then every nonzero element a R has a factorization into irreducible
factors (and a unit).
Proof. We may assume a = a
0
is not a unit. Thus by Lemma 3.9, a has some irreducible factor p
1
; say
a = p
1
a
1
. If a
1
is a unit were done, otherwise, we can write a
1
= p
2
a
2
for some irreducible p
2
. Continuing
by induction, if we never get a unit, we get an innite sequence . . . a
2
|a
1
|a
0
of nontrivial factors, which gives
a contradiction as in the proof of Lemma 3.9. Thus some a
n
is eventually a unit, so we get a factorization
a = a
n
p
1
p
n
.
Corollary 3.11. Let R be a PID. Then R is a UFD.
3.1 Exercises
Exercise 3.1. Let k be a eld. For f(x, y) k[x, y], let deg(f) be the largest degree of a term of f, where
the degree of a monomial x
m
y
n
is dened to be m + n. Show that deg does not make k[x, y] a Euclidean
domain.
Exercise 3.2. Show that Z[

2] is a Euclidean domain, and thus a PID and a UFD. (Hint: Imitate the
proof for Z[i].)
I dont know how to do the following exercise, but I know that its true. Id love to see a solution if you
can nd one. The diculty is that you cant use the geometry of complex numbers, so we cant make exactly
the same sort of argument as we did for Z[i].
Exercise 3.3. Show that Z[

2] is a Euclidean domain. (Hint: I think d(a + b

2) = a
2
+ 2b
2
works.)
Exercise 3.4. Call a Euclidean domain R strictly Euclidean if whenever neither a nor b is a unit, then
d(ab) > max(d(a), d(b)). For example, Z, k[x], and Z[i] are all strictly Euclidean. Show using induction on
d(a) (instead of the argument using ideals given in class) that if R is strictly Euclidean, then any nonzero
a R can be factored into irreducibles.
Exercise 3.5. Let R be a domain. Then the eld of fractions K(R) of R is the collection of expressions
a/b with a, b R, b = 0, where we say a/b = c/d if ad = bc. We make K(R) a ring using the ordinary rules
for addition and multiplication of fractions.
(a): Show that by mapping a R to a/1, we can think of R as a subring of K.
(b): Suppose R is a UFD. Show that any nonzero element of K(R) can be factored as u

p
di
i
where u R
is a unit, p
i
R are irreducible, and d
i
Z are possibly negative, and this factorization is unique up to
changing the factors by units in R.
8
4 Factorization in polynomial rings
In the previous section, we showed that if k is a eld, then the ring of polynomials k[x] is a UFD. In this
section we will look at polynomials with more general coecients. Our basic example will be Z[x], the ring
of polynomials with integer coecients.
We can start by trying to use the same method we did in the case of a eld: we want to show that Z[x]
is a Euclidean domain with d(f) = deg(f). The main thing to check is that given any f and g, we can write
g = qf + r where the remainder r has lower degree than f. Over a eld, we can do this by polynomial long
division.
However, over Z this breaks down. Consider, for example, g(x) = x
2
+x+2 and f(x) = 2x+1. Performing
long division would give quotient
1
2
x +
1
4
with remainder 7/4. However, were not allowed to do this since
the coecients must remain integers. If the quotient has to have integer coecients, theres no way we can
use a multiple of f to cancel out the x
2
and x terms in g and leave an integer remainder.
Indeed, it is impossible to make Z[x] a Euclidean domain since it is not a PID! For example, the ideal
(2, x) is not principal: it is the set of polynomials with even constant term, which is not just the multiples
of any one polynomial. Alternatively, although clearly the gcd of 2 and x should be 1, there is no way to
write 1 as a linear combination of 2 and x.
Nevertheless, it turns out that Z[x] is a UFD. To show this, we will need a new argument dierent from
the ones weve used. The basic idea will be to use the fact that both Z and Q[x] are UFDs, and the fact
that Z[x] is kind of halfway between them.
Lets look at what irreducibles in Z[x] look like. There are two kinds, namely the primes in Z as constant
polynomials, and the irreducible polynomials of degree > 0. In particular, note that if f(x) Z[x] is
irreducible as an element of Q[x] and monic, it is also automatically irreducible in Z[x], since there are only
fewer possible factorizations in Z[x] (and since it is monic, there is no issue with the fact that Z[x] has fewer
units). Thus the rst kind of irreducibles are the irreducibles of Z, and the second kind at least include some
irreducibles coming from Q[x].
To show Z[x] is a UFD, the main thing we need to show is that all irreducibles are prime, so we will start
by showing this for the irreducibles coming from Z. In fact, nothing we will do is special to Z, so we will
replace it with an arbitrary UFD.
Lemma 4.1 (Gausss Lemma). Let R be a domain and let p R be prime (e.g., R is a UFD and p is
irreducible). Then p is also prime as an element of R[x].
Proof. Lets see what it means for p to be prime in R[x]. Note that p divides a polynomial i it divides each
of its coecients. Thus the claim is that if p divides all the coecients of fg, it must divide either all the
coecients of f or all the coecients of g.
This is far from obvious, since the coecients of fg are complicated expressions in the coecients of f
and g. Lets look at an example: say f(x) = ax + b and g(x) = cx + d. Then
f(x)g(x) = acx
2
+ (ad + bc)x + bd.
Lets suppose all the coecients of the product are divisible by p. Then in particular, p|ac, so p|a or p|c.
Say p|a (the other case is similar). Then since p|ad + bc, the next coecient, we must have p|bc so p|b or
p|c. If p|b, then p|ax + b = f and were done. If p |b, then p|c, and also since p|bd, the next coecient, p|d.
Thus p|cx + d = g, and were done again.
The proof for general polynomials is similar to this example. Say deg(f) = n and deg(g) = m and write
f(x) = a
n
x
n
+ + a
0
and
g(x) = b
m
x
m
+ + b
0
.
Lets assume that p |f; we will show p|g. There is some coecient of f that p does not divide; let a
r
be the
largest one (i.e. the one with r maximal). We show by descending induction that p divides each coecient
9
of g. First, the coecient of x
m+r
in f(x)g(x) will be
b
m
a
r
+ b
m1
a
r+1
+ b
m2
a
r+2
+ . . .
By maximality of r, p divides a
r+1
, a
r+2
, and so on, so p divides every term but the rst. But p|fg so p
must divide the entire coecient, so p|b
m
a
r
. Since p |a
r
, p|b
m
.
Now let k < m and suppose p|b

for all > k. Then the coecient of x

k+r
in f(x)g(x) is
+ b
k+1
a
r1
+ b
k
a
r
+ b
k1
a
r+1
+ . . .
Since p|b

for > k and p|a

s
for s > r, we know that p divides every term of this except b
k
a
r
. Thus p|b
k
a
r
and so p|b
k
.
Gausss lemma not only tells us that the irreducibles from Z are still prime in Z[x]; it also tells us
something about the nonconstant irreducibles. For this, we rst need a denition.
Denition 4.2. Let R be a UFD and f R[x]. Then the content cont(f) of f (dened up to a unit) is the
gcd of all the coecients of f (gcd makes sense because R is a UFD). If cont(f) = 1, we say f is primitive.
Note that for any f, we can factor f = cont(f)f
1
where f
1
is primitive.
Corollary 4.3. Let R be a UFD and f, g R[x]. Then cont(fg) = cont(f) cont(g).
Proof. By factoring f = cont(f)f
1
and g = cont(g)g
1
, we get fg = cont(f) cont(g)f
1
g
1
. It thus suces to
show that f
1
g
1
is primitive. If it were not primitive, some irreducible p R would divide it. But by Gausss
Lemma, this would imply p|f
1
or p|g
1
, which is impossible since they are primitive.
Corollary 4.4. A nonconstant element of Z[x] is irreducible i it is primitive and is irreducible as an
element of Q[x].
Proof. First, suppose f Z[x] is irreducible and nonconstant. It clearly must be primitive. Suppose we
can factor f = gh for g, h Q[x]. Note that content also makese sense for elements of Q[x]: it is the
unique rational cont(g) (up to a unit in Z such that g = cont(g)g
1
for some primitive g
1
Z[x] (explicitly,
write all the coecients of g in lowest terms, and take the gcd of the numerators divided by the lcm of
the denominators). We can thus write f = cont(g) cont(h)g
1
h
1
, where now g
1
, h
1
Z[x] are primitive
and cont(g), cont(h) Q. By Corollary 4.3, g
1
h
1
is primitive as well. But f is primitive, so this implies
cont(g) cont(h) must be a unit in Z. But then, up to a unit, we can factor f = g
1
h
1
in Z[x]. Since f is
irreducible in Z[x], this implies g
1
or h
1
is a unit, so then g or h is a unit in Q[x]. Thus f is irreducible in
Q[x].
Conversely, suppose f is primitive and is irreducible as an element of Q[x], and suppose we can factor
f = gh in Z[x]. Since f is irreducible in Q[x], g or h must be a unit in Q[x], i.e. an element of Q (or actually
Z since it is also in Z[x]). But since f is primitive, f has no nontrivial factors in Z, so g or h must be a unit
in Z. Thus f is irreducible in Z[x].
Using this, we can nally prove:
Theorem 4.5. Z[x] is a UFD.
Proof. First, we show the existence of factorizations. Let f Z[x] and factor f =

q
i
where the q
i
are
irreducible elements of Q[x] (this is possible since Q[x] is a UFD). By factoring out the contents of the q
i
, we
can rewrite this as f = c

i
, where q

i
Z[x] is primitive and irreducible in Q[x] and c Q is the product
of the contents of the q
i
. By Corollary 4.4, the q

i
are irreducible in Z[x], and by Corollary 4.3, c is equal
to the content of f and is thus actually in Z. Factoring c into primes

p
j
in Z then gives a factorization
f =

p
j

i
of f into irreducibles of Z[x].
We now show that every irreducible in Z[x] is prime. By Gausss lemma, it suces to show this for
nonconstant irreducibles f. Suppose f|gh. By Corollary 4.4, f is irreducible in Q[x] and hence prime in Q[x]
10
since Q[x] is a UFD. Thus in Q[x], f|g or f|h; say f|g. We can then write g = qf, where q Q[x]. Now we
can write q = cont(q)q
1
where cont(q) Q and q
1
Z[x] is primitive. Since f is primitive, it follows that
cont(q) = cont(g) is actually an integer, so in fact q = cont(q)q
1
is actually in Z[x]. Thus f|g in Z[x]. and f
is prime.
Finally, we observe that in all of this, there was nothing special about Z: we could have replaced Z by
any UFD R, as long as we also replace Q by the eld of fractions of R. Recall that the eld of fractions of a
domain R is the set of formal fractions a/b with a, b R, b = 0; this is a eld with the usual multiplication
and addition of fractions and R embeds in it by sending a to a/1.
Theorem 4.6. Let R be a UFD. Then R[x] is a UFD.
Proof. Exactly the same as the proof of Corollary 4.4 and Theorem 4.5, with Z replaced by R and Q replaced
by its eld of fractions.
In particular, by induction we get that a polynomial ring k[x, y, z, . . . ] or Z[x, y, z, . . . ] in any (nite)
number of variables is a UFD. Furthermore, we can use Corollary 4.4 to get an inductive criterion for
detecting irreducibility of elements of these polynomial rings. However, this criterion is hard to use since
the elds of fractions that come up in the induction steps, such as the eld of fractions of k[x, y], are not
particularly easy to work with.
One useful application of unique factorization is showing the quotients are domains. By Exercise 2.6, if
R is a domain and p R is prime, then the quotient R/(p) is a domain. In general, it can be quite dicult
to show that a particular ring dened as a quotient is a domain. However, for UFDs, we get that R/(p) is a
domain for any irreducible p. In particular, whenever you have an irreducible polynomial f k[x, y, z, . . . ],
the quotient k[x, y, z, . . . ]/(f) is a domain.
4.1 Exercises
Exercise 4.1 (Eisensteins criterion). Let R be a UFD, p R be prime, and let f(x) = a
n
x
n
+a
n1
x
n1
+
+a
0
R[x] be a polynomial. Suppose that p|a
i
for all i < n, p |a
n
and p
2
|a
0
. Show that f is irreducible.
Exercise 4.2. Let k be a eld. Show that k[x, y, z, w]/(xy zw) is a domain but not a UFD. (Hint: See
Exercises 1.5 and 2.6.)
Exercise 4.3. Let R be a ring that is not a eld. Show that R[x] is not a PID.
Exercise 4.4. Let R be a UFD and suppose d R is not a square (i.e., d = a
2
for any a R).
(a): Show that x
2
d R[x] is irreducible.
(b): Suppose R = Z. Show that Z[x]/(x
2
d) is isomorphic to Z[

d]. (This gives a new proof that Z[

d]
is a domain, which we originally showed by saying it is a subring of C.)
11
5 Dedekind domains
Lets look more closely at the case Z[

3]. By trial and error, we can nd that unique factorization fails:

for example, 2 2 = (1+

3)(1

3). However, it turns out that in this case there is a more general reason
why unique factorization must fail. To see this, we rst need some denitions.
Denition 5.1. Let S be a ring and let R S be a subring, and let a S. We say a is integral over R if
there is a monic polynomial f(x) = x
n
+c
n1
x
n1
+ +c
0
R[x] such that f(a) = 0. We say a domain R
is integrally closed if whenever a K(R), the eld of fractions of R, is integral over R, then in fact a R.
Note that, for example, any a R is integral over R, by letting f(x) = x a. On the other hand, we
drop the requirement that f be monic, then all of K(R) is integral over R, since a fraction a/b is a root of
f(x) = bx a.
In fact, we already know some examples of integrally closed domains.
Proposition 5.2. UFDs are integrally closed.
Proof. Let R be a UFD and let a/b K(R) be integral over R; say
(a/b)
n
+ c
n1
(a/b)
n1
+ + c
0
= 0.
We assume a and b have no common factor (so a/b is in lowest terms). We can then multiply the equation
above by b
n
to obtain
a
n
+ bc
n1
a
n1
+ + b
n
c
0
= 0,
or
a
n
= bc
n1
a
n1
b
n
c
0
.
But the right-hand side is divisible by b, and a and b had no common factor and hence since R is a UFD, a
n
and b have no common factor. The only way this is possible is if b is a unit. But in that case, a/b is actually
in R, so R is integrally closed.
On the other hand, we also have a non-example:
Proposition 5.3. Z[

3] is not integrally closed.

Proof. Let = (1+

3)/2. Then is certainly in the eld of fractions of Z[

3], and it is not in Z[

3].
Thus if we show is integral, Z[

3] will not be integrally closed. To show this, we can simply note that
is a root of f(x) = x
2
+ x + 1, by the quadratic formula.
Remark 5.4. Note that more generally, (1 +

d/2 is a root of x
2
+x+(d +1)/4, so whenever d 3 mod
4 this shows that Z[

d] is not integrally closed.

Thus Z[

3] in some sense lacks unique factorization for a stupid reason that is easily remedied: its
just missing some elements it should have. More precisely, we have the following fact, whose proof is beyond
the scope of this class.
Theorem 5.5. Let R be a domain. Then the set R of elements of K(R) integral over R is a ring and is
integrally closed, called the integral closure of R.
That is, if we just take R and throw in all the elements that are integral over it, we can complete R
to an integrally closed domain, which now has some hope of being a UFD. Indeed, this works in our

3
example.
Theorem 5.6. Let = (1 +

3)/2 as above. Then Z[] is the integral closure of Z[

3], and is a UFD.

12
Proof. If Z[] is a UFD, it is automatically the integral closure of Z[

3]: indeed, it is integrally closed,

and therefore contains the integral closure of Z[

3]. But is integral over Z[

3], so it is also contained

in the integral closure of Z[

3].
We thus show that Z[] is a UFD. We use the same argument as we always have: we show that it is a
Euclidean domain under the norm function given by the square of the complex absolute value, by a geometric
argument. Now instead of a rectangle we have a parallelogram given by 1 and in the complex plane, but
the geometric computation still works out.
A similar argument shows that Z[(1 +

7)/2] is a UFD and the integral closure of Z[

7]. More
generally, it is always true that for d 3 mod 4 (and squarefree), Z[(1 +

d)/2] is the integral closure

of Z[

d]. However, the geometric arguments quickly break down as d gets large, and in fact we have the
following theorem.
Theorem 5.7. Let d > 0 be squarefree. Then the integral closure of Z[

d] is a UFD i d is 1, 2, 3, 7,
11, 19, 43, 67, or 163.
This result is far beyond the scope of this class, and was proven only in 1952.
Nevertheless, it turns out that there is a generalization of unique factorization that often works in domains
that are not UFDs. The idea is to not attempt to factorize elements of the ring, but ideals. This is a natural
thing to do because the ideal (a) generated by an element a depends on a only up to multiplication by units
(and in fact determines a, up to a unit), and in a PID ideals are exactly the same as elements, up to units.
Indeed, this was historically the original reasons that ideals were dened: they are ideal elements that can
be substitutes for elements (up to units) for purposes of factorization, but are better-behaved in domains
that are not PIDs.
To make sense of this, we should rst dene multiplication of ideals.
Denition 5.8. Let I, J R be ideals. Then IJ is the ideal of elements {

c
n
d
n
: c
n
I, d
n
J}.
Note that in the case of principal ideals, we have (a)(b) = (ab), since every term of the sum

c
n
d
n
is
divisible by ab, so the whole sum is. More generally, if I is generated by {a
i
} and J is generated by {b
j
},
then IJ will be generated by {a
i
b
j
}.
We also have a denition of a prime ideal, which we have already seen in Exercise 2.6.
Denition 5.9. An ideal I R is prime if whenever ab I, then either a I or b I.
Note that a principal ideal (p) is prime i the element p is prime.
Lets see an example how using prime ideals instead of irreducible elements might be able to save unique
factorization.
Example 5.10. Recall that in Z[

5], 9 has two dierent irreducible factorizations, 3 3 and (2+

5)(2

5). However, these do not give two dierent factorizations of the ideal (9), since the ideals (3) and
(2

5) are not prime! Indeed, since 3|9 but 3 |2

5, (3) cannot be prime, and similarly for (2

5.
However, it turns out that the non-principal ideals I = (3, 2 +

5) and J = (3, 2

5) are prime.
Furthermore, we have
IJ = (3, 2 +

5) (3, 2

5) = (9, 6 3

5, 6 + 3

5, 9).
Note that everything on the right-hand side is divisible by 3, so IJ (3). On the other hand, since 6+3

5
and 6 3

5 are both in IJ, so is their dierence 12, so IJ (9, 12) = (3). Thus IJ = (3). A similar
computation shows that I
2
= (2 +

5) and J
2
= (2

5). Thus if we think in terms of ideals, the failed

uniqueness of factorization 3 3 = (2 +

5)(2

5) becomes just IJ IJ = I
2
J
2
, both giving the same
factorization of (9) into prime ideals.
However, using prime ideals will not always save us. As an example, consider the ring k[t
2
, t
3
], which we
earlier saw was not a UFD because t
2
and t
3
are both irreducible but (t
2
)
3
= (t
3
)
2
. In this case, the only
13
prime ideal containing either (t
2
) or (t
3
) is the ideal I = (t
2
, t
3
). However, I
2
= (t
4
, t
5
, t
6
) = (t
4
) is already
smaller than both (t
2
) and (t
3
). This means that (t
2
) and (t
3
) cannot even be written as a product of prime
ideals at all, let alone uniquely. It turns out the problem is that t
3
/t
2
= t should be an element of this ring,
so that instead of I = (t
2
, t
3
) we could have I = (t), and then everything would work. More precisely, this
ring is not integrally closed, since t is a root of f(x) = x
2
t
2
and hence integral.
It turns out that being integrally closed is almost all you need to be able to use ideals to get unique
factorization.
Theorem 5.11. Let R be an integrally closed domain which is Noetherian and in which every prime ideal
except {0} is maximal (a ring satisfying these hypotheses is called a Dedekind domain. Then every ideal
I R factors uniquely as a product

P
di
i
where the P
i
are prime ideals.
Again, it would take more time (or more chilies) than we have to prove this theorem. Here Noetherian is
a technical condition that says that R is not too large and pathological (more precisely, a ring is Noetherian
if every ideal can be generated by a nite set of elements). The other condition, that every prime ideal is
maximal, is satised by any PID and by the integral closure of Z[

d] for any d, and it turns out that this

condition together with being a UFD actually implies that your ring is a PID. However, we have already
seen that there are UFDs that are not PIDs and hence do not satisfy this property, namely polynomial rings
over UFDs that are not elds.
5.1 Exercises
For the exercises, you need the following result, which you can try to prove if youre interested (though it is
hard without using tools we have not introduced in this class).
Theorem 5.12. Let R S be domains a S be integral over aR, and let f(x) R[x] be of minimal degree
such that f(a) = 0. Then the leading coecient of f is a unit in R. (The denition of integral we gave
before only stipulates that there is some monic polynomial f such that f(a) = 0.)
Exercise 5.1. Show that if d is square-free and d 3 mod 4, then Z[(1+

d)/2] is integrally closed. (Hint:

The eld of fractions is Q[

d]. Given an element z = a +b

d with a, b Q, nd a quadratic polynomial

that z is a root of, and show this polynomial can be made a monic polynomial in Z[x] i z Z[(1+

d)/2].
Conclude that Z[(1 +

d)/2] is the integral closure of Z in Q[

d], and hence is integrally closed by

Theorem 5.5.)
Exercise 5.2. Show that if d is square-free and d 1 mod 4, then Z[

d] is integrally closed. (Hint: This

is similar to the previous exercise.)
Exercise 5.3. Show that Z[(1 +

7)/2] is a Euclidean domain, and hence a PID and UFD.

14