Integral Domains

As always in this course, a ring R is understood to be a commutative

ring with unity.

1 First definitions and properties

Definition 1.1. Let R be a ring. A divisor of zero or zero divisor in R is
an element r ∈ R, such that there exists an s ∈ R with s 6= 0 and rs = 0.
Thus, for example, 0 is always a zero divisor.

Example: in Z/6Z, 0 = 2 · 3, hence both 2 and 3 are divisors of zero.

One way to find divisors of zero is as follows:

Definition 1.2. Let R be a ring. A nilpotent element of R is an element r,

such that there exists an n ∈ N such that rn = 0. Note that 0 is allowed to
be nilpotent.

Lemma 1.3. Let R be a ring and let r ∈ R be nilpotent. Then r is a zero

divisor.

Proof. The set of n ∈ N such that rn = 0 is nonempty, so let m be the

smallest such natural number. Note that, if m = 1, r = 0 and hence is a
divisor of zero. Otherwise, 0 < m − 1 < m, so by assumption rm−1 6= 0.
Hence r · rm−1 = rm = 0, with rm−1 6= 0, so that r is a divisor of zero.

Example: in Z/16Z, 0 = 24 = 2 · 23 , hence 2 is a divisor of zero. But in

Z/6Z, neither 2 nor 3 is nilpotent, so there are examples of divisors of zero
which are not nilpotent.

Definition 1.4. A ring R is an integral domain if R 6= {0}, or equivalently

1 6= 0, and such that r is a zero divisor in R ⇐⇒ r = 0. Equivalently,
a nonzero ring R is an integral domain ⇐⇒ for all r, s ∈ R with r 6= 0,
s 6= 0, the product rs 6= 0 ⇐⇒ for all r, s ∈ R, if rs = 0, then either r = 0
or s = 0.

1
Definition 1.5. Let R be a ring. The cancellation law holds in R if, for all
r, s, t ∈ R such that t 6= 0, if tr = ts, then r = s.
Lemma 1.6. A ring R 6= {0} is an integral domain ⇐⇒ the cancellation
law holds in R.
Proof. =⇒ : if tr = ts and t 6= 0, then tr − ts = t(r − s) = 0. Since t 6= 0
and R is an integral domain, r − s = 0 so that r = s.
⇐= : Suppose that rs = 0. We must show that either r or s is 0. If
r 6= 0, then apply cancellation to rs = 0 = r0 to conclude that s = 0.

The following are examples of integral domains:

1. A field is an integral domain. In fact, if F is a field, r, s ∈ F with
r 6= 0 and rs = 0, then 0 = r−1 0 = r−1 (rs) = (r−1 r)s = 1s = s.
Hence s = 0. (Recall that 1 6= 0 in a field, so the condition that F 6= 0
is automatic.) This argument also shows that, in any ring R 6= 0, a
unit is not a zero divisor.

2. If S is an integral domain and R ≤ S, then R is an integral domain.

In particular, a subring of a field is an integral domain. (Note that,
if R ≤ S and 1 6= 0 in S, then 1 6= 0 in R.) Examples: √ any √subring
of R or C is an integral domain. Thus for example Z[ 2], Q( 2) are
integral domains.

3. For n ∈ N, the ring Z/nZ is an integral domain ⇐⇒ n is prime. In

fact, we have already seen that Z/pZ = Fp is a field, hence an integral
domain. Conversely, if n is not prime, say n = ab with a, b ∈ N, then,
as elements of Z/nZ, a 6= 0, b 6= 0, but ab = n = 0. Hence Z/nZ is
not an integral domain.

4. If R is an integral domain, then, as we shall see in a minute, R[x]

is an integral domain. Hence, by induction, R[x1 , . . . , xn ] is an inte-
gral domain if R is an integral domain. In particular, if F is a field,
F [x1 , . . . , xn ] is an integral domain, as is Z[x1 , . . . , xn ].
To prove the last statement (4) above, we show in fact:
Lemma 1.7. Let R be an integral domain. Then, if f, g ∈ R[x] are both
nonzero, then f g 6= 0 and deg(f g) = deg f + deg g.
Proof. Let d = deg f and e = deg g. Then f = di=0 ai xi and g = ej=0 bj xj
P P

with ad , be 6= 0. Since ad be 6= 0, the leading term of f g is ad be xd+e . Hence

f g 6= 0 and deg(f g) = d + e = deg f + deg g.

2
Corollary 1.8. Let R be an integral domain. Then the group of units
(R[x])∗ in the polynomial ring R[x] is just the group of units R∗ in R (viewed
as constant polynomials).
Proof. Clearly, if u is a unit in R, then it is a unit in R[x], so that R∗ ⊆
(R[x])∗ . Conversely, if f ∈ (R[x])∗ , then there exists a g ∈ R[x] such that
f g = 1. Clearly, neither f nor g is the zero polynomial, and hence

0 = deg 1 = deg(f g) = deg f + deg g.

Thus, deg f = deg g = 0, so that f , g are elements of R and clearly they

are units in R. Hence f ∈ R∗ , so that (R[x])∗ ⊆ R∗ . It follows that
(R[x])∗ = R∗ .

The corollary fails if the ring R has nonzero nilpotent elements. For
example, in (Z/4Z)[x],

(1 + 2x)(1 + 2x) = (1 + 2x)2 = 1 + 4x + 4x2 = 1,

so that 1 + 2x is a unit in (Z/4Z)[x].

Finally, we note the following:
Proposition 1.9. A finite integral domain R is a field.
Proof. Suppose r ∈ R with r 6= 0. The elements 1 = r0 , r, r2 , . . . cannot all
be different, since otherwise R would be infinite. Hence there exist 0 ≤ n <
m with rn = rm . Writing m = n + k with k ≥ 1, we see that rn = rm =
rn+k = rn rk . By induction, since R is an integral domain and r 6= 0, rn 6= 0
for all n ≥ 0. Applying cancellation to rn = rn · 1 = rn rk gives rk = 1.
Finally since rk = r · rk−1 , we see that r is invertible, with r−1 = rk−1 .

2 The characteristic of an integral domain

Let R be an integral domain. As we have seen in the homework, the function
f : Z → R defined by f (n) = n · 1 is a ring homomorphism and its image is
h1i, the cyclic subgroup of (R, +) generated by 1. There are two possibilities:
(1) 1 has finite order n, in which case h1i ∼
= Z/nZ, or (2) 1 has infinite order,
∼
in which case h1i = Z.
Proposition 2.1. With notation as above,
(i) If 1 has finite order n, then n = p is a prime number, and every
nonzero element of R has order p.

3
 (ii) If 1 has infinite order, then every nonzero element of R has infinite
order.

Proof. (i) By definition, n is the smallest positive integer such that n · 1 = 0.

If n = ab, where a, b ∈ N, then (using homework) 0 = n · 1 = (a · 1)(b · 1).
Since R is an integral domain, one of a·1, b·1 is 0. Say a·1 = 0. Then a ≥ n,
but since a divides n, we must have a = n. Hence in every factorization
of n, one of the factors is n, so by definition n is a prime p. Moreover, for
every r ∈ R, p · r = (p · 1)r = 0, so that the order of r divides p. If r 6= 0,
then its order is greater than 1, hence must equal p.
(ii) Let r ∈ R, and suppose that r has (finite) order n ∈ N, so that
n · r = 0. As in the proof of (i), write n · r = (n · 1)r. Since 1 has infinite
order, n · 1 6= 0, and hence r = 0. Thus, if r 6= 0, then n · r 6= 0 for every
n ∈ N. Thus r has infinite order.

Definition 2.2. Let R be an integral domain. If 1 ∈ R has infinite order, we

say that the characteristic of R is zero. If 1 ∈ R has finite order, necessarily
a prime p, we say that the characteristic of R is p. In either case we write
char R for the characteristic of R, so that char R is either 0 or a prime
number.

Examples: Clearly, the characteristic of Z is 0. Also, if R and S are

integral domains with R √ ≤ S, then clearly char R = char S. Thus char Q,
char R, char C, char Q( 2), etc. are all 0. On the other hand, the charac-
teristic of Fp = Z/pZ is p. Thus, the characteristic of Fp [x] is also p, so that
Fp [x] is an example of an infinite integral domain with characteristic p 6= 0,
and Fp [x] is not a field. (Note however that a finite integral domain, which
automatically has positive characteristic, is always a field.)

3 The field of quotients of an integral domain

We first begin with some general remarks about fields. If F is a field and
r, s ∈ F with s 6= 0, we write (as usual) rs−1 = r/s. Note that r/s = t/w
⇐⇒ rw = st, since rw = (sw)r/s and st = (sw)t/w, and by cancellation.
Then the laws for adding and multiplying fractions are forced by associativ-
ity and distributivity in F : for example,

r/s + t/w = rs−1 + tw−1 = (rw)(sw)−1 + (ts)(sw)−1

= (rw + ts)(sw)−1 = (rw + ts)/(sw).

4
 Now suppose that R is an integral domain. We would like to enlarge R
to a field, in much the same way that we enlarge Z to Q. To this end, we
construct a set whose elements are “fractions” r/s with r, s ∈ R and s 6= 0.
Two fractions r/s and t/w are identified if, as in the discussion above for
fields, rw = st. The correct way to say this is via equivalence classes: on
the set R × (R − {0}), define the relation ∼ on pairs (r, s) by: (r, s) ∼ (t, w)
⇐⇒ rw = st.
Lemma 3.1. ∼ is an equivalence relation.
Proof. We must show ∼ is reflexive, symmetric, and transitive. Reflex-
ive: (r, s) ∼ (r, s) ⇐⇒ rs = sr, which holds since R is commutative.
Symmetric: (r, s) ∼ (t, w) ⇐⇒ rw = st, in which case ts = wr, hence
(t, w) ∼ (r, s). Transitive (it is here that we use the fact that R is an in-
tegral domain): suppose that (r, s) ∼ (t, w) and that (t, w) ∼ (u, v), with
s, w, v 6= 0. By definiton rw = st and tv = wu. Then rwv = stv = swu,
hence w(rv) = w(su). Since w 6= 0 and R is an integral domain, rv = su,
hence (r, s) ∼ (u, v). Thus ∼ is transitive.

Define Q(R), the field of quotients of R, to be the set of equivalence

classes (R × (R − {0}))/ ∼. Next we need operations of addition and mul-
tiplication on Q(R). As is usually the case with equivalence relations, we
define these operations by defining them on representative of equivalence
classes, and then check that the operations are in fact well-defined. Define

[(r, s)] + [(t, w)] = [(rw + st, sw)]; [(r, s)] · [(t, w)] = [(rt, sw)].

Lemma 3.2. Let ∼ and Q(R) be as above.

(i) The operations of addition and multiplication are well-defined.

(ii) (Q(R), +, ·) is a field.

(iii) The function ρ : R → Q(R) defined by ρ(r) = [(r, 1)] is an injective

homomorphism.
Proof. These are all straightforward if sometimes tedious calculations. For
example, to see (i), suppose that (r, s) ∼ (r0 , s0 ). We shall show that (rw +
st, sw) ∼ (r0 w + s0 t, s0 w) and that (rt, sw) ∼ (r0 t, s0 w). By definition, rs0 =
sr0 . Then

(rw + st)(s0 w) = rws0 w + sts0 w = (rs0 )(w2 ) + (ss0 )(tw)

= (r0 s)(w2 ) + (ss0 )(tw) = (r0 w + s0 t)(sw).

5
Hence (rw + st, sw) ∼ (r0 w + s0 t, s0 w). Moreover,

(rt)(s0 w) = (rs0 )(tw) = (r0 s)(tw) = (r0 t)(sw).

Hence (rt, sw) ∼ (r0 t, s0 w). Similarly, if (t, w) ∼ (t0 , w0 ), then (rw+st, sw) ∼
(rw0 + st0 , sw0 ) and that (rt, sw) ∼ (rt0 , sw0 ).
To see (ii), we must show first that (Q(R), +) is an abelian group and that
multiplication is associative, commutative, and distributes over addition.
These are all completely straightforward if long computations. Note that
[(0, 1)] = [(0, r)] is the additive identity, that [(r, s)] ∼ [(0, 1)] ⇐⇒ r = 0,
and that [(1, 1)] = [(r, r)] is a multiplicative identity. Finally, if [(r, s)] 6=
[(0, 1)], so that r 6= 0, then [(s, r)] ∈ Q(R) and [(r, s)][(s, r)] = [(rs, rs)] =
[(1, 1)]. Thus Q(R) is a field.
To see (iii), defining ρ(r) = [(r, 1)], we see that

ρ(r + s) = [(r + s, 1)] = [(r, 1)] + [(s, 1)] = ρ(r) + ρ(s);

ρ(rs) = [(rs, 1)] = [(r, 1)][(s, 1)] = ρ(r)ρ(s).

Thus ρ is a homomorphism. It is injective since ρ(r) = ρ(s) ⇐⇒ (r, 1) ∼

(s, 1) ⇐⇒ r = s.

From now on we write [(r, s)] as r/s or as rs−1 and identify r ∈ R with
its image r/1 ∈ Q(R). In this way we view R as a subring of Q(R).
Example: 1) let F be a field and F [x] the polynomial ring with coeffi-
cients in F . Then we denote Q(F [x]) by F (x). By definition, the elements
of F (x) are quotients f /g, where f , g are polynomials with coefficients in
F . We call F (x) the field of rational functions with coefficients in F . In
particular, taking F = Fp , the field of rational functions Fp (x) is an example
of an infinite field (since it contains a subring isomorphic to the polynomial
ring Fp [x], which is infinite), whose characteristic is p > 0.
2) If R = F is already a field, then (r, s) ∼ (rs−1 , 1). Thus the injective
homomorphism ρ is also surjective, hence an isomorphism, so that Q(F ) ∼ =
F.
Remark: In the field of quotients Q = Q(Z) of Z, we can always put a
fraction n/m in lowest terms, i.e. we can assume that gcd(n, m) = 1. This
says that the equivalence class [(n, m)] has a “best” representative, if we
require in addition, say, that m > 0. Such a choice depends on results about
factorization in Z, and is not possible in a general integral domain.
Finally, we show that Q(R) has a very general property with respect to
injective homomorphisms from R to a field:

6
Proposition 3.3. Let R be an integral domain, F a field, and φ : R → F be
an injective homomorphism. Then there exists a unique injective homomor-
phism φ̃ : Q(R) → F such that φ̃(r/1) = φ(r). Finally, if every element of
F is of the form φ(r)/φ(s) for some r, s ∈ R with s 6= 0, then φ̃ : Q(R) → F
is an isomorphism, and in particular Q(R) ∼ = F.

Proof. Clearly, if φ̃ exists, then we must have

φ̃(r/s) = φ̃(rs−1 ) = φ̃(r)φ̃(s−1 ) = φ̃(r)φ̃(s)−1 = φ̃(r)/φ̃(s) = φ(r)/φ(s).

(Here we have used the fact, which is easy to check, that φ̃(s−1 ) = φ̃(s)−1 .)
This proves that φ̃ is unique, if it exists. Conversely, we try to define φ̃ by
the formula
φ̃(r/s) = φ(r)/φ(s).
Here r/s is shorthand for the equivalence class [(r, s)] ∈ Q(R), and the
fraction φ(r)/φ(s) = φ(r)/φ(s)−1 is well-defined in F since, as φ is injec-
tive and s 6= 0, φ(s) 6= 0. We must first show that φ̃ is well-defined, i.e.
independent of the choice of representative (r, s) ∈ [(r, s)]. Choosing an-
other representative (r0 , s0 ) ∈ [(r, s)], we have by definition rs0 = r0 s. Hence
φ(rs0 ) = φ(r)φ(s0 ) = φ(r0 s) = φ(r0 )φ(s). Dividing by φ(s)φ(s0 ) gives

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

Hence φ̃(r/s) = φ(r)/φ(s) is independent of the choice of representative

(r, s) ∈ [(r, s)]. It is then straightforward to check that φ̃ is a (ring) iso-
morphism. To see that it is injective, suppose that φ̃(r/s) = φ̃(r0 /s0 ). Then
φ(r)/φ(s) = φ(r0 )/φ(s0 ), and hence

φ(rs0 ) = φ(r)φ(s0 ) = φ(r0 )φ(s) = φ(r0 s).

Since φ is injective, rs0 = r0 s, and hence r/s = r0 /s0 . Thus φ̃ is injective. (In
fact, this is a general property of ring homomorphisms where the domain
is a field.) Finally, if every element of F is of the form φ(r)/φ(s) for some
r, s ∈ R with s 6= 0, then φ̃ is also surjective, hence an isomorphism.

Here is a typical way we might apply the proposition:

Lemma 3.4. Let R be an integral domain with field of quotients Q(R). Then
Q(R[x]), the field of quotients of the integral domain R[x], is isomorphic to
Q(R)(x), the field of rational functions with coefficients in Q(R).

7
Proof. Since R is isomorphic to a subring of Q(R), there is a natural ho-
momorphism from R[x] to Q(R)[x], and since Q(R)[x] is isomorphic to a
subring of its field of quotients Q(R)(x), there is an injective homomor-
phism from R[x] to Q(R)(x), which amounts to viewing a polynomial with
coefficients in R as a particular example of a rational function with co-
efficients in Q(R). Hence, by the proposition, there is an injective ho-
momorphism Q(R[x]) → Q(R)(x). To see that it is surjective, it suffices
to show that every rational function with coefficients in Q(R) is a quo-
tient of two polynomials P with coefficients P in R. Given such a quotient
f /g, suppose that f = ni=0 ai xi and g = m j
j=0 bj x , with ai , bj ∈ Q(R).
Then ai = ri /si with ri , si ∈ R and si 6= 0. Likewise, bj = tj /wj with
tj , wj ∈ R and wj 6= 0. We then proceed to “clear Q denominators” in the
coefficients: LetQN = s0Q· · · · · sn · w0 · · · · · wm = ni=0 si · m
Q
j=0 j Then
w .
m
N (rk /sk ) = rk i6=k si · j=0 wj ∈ R, and similarly N (tj /wj ) ∈ R. Clearly
N f ∈ R[x] and N g ∈ R[x]. Thus

f f N Nf
= · = .
g g N Ng

It then follows that f /g = N f /N g is a quotient of two polynomials with

coefficients in R. Hence Q(R[x]) ∼
= Q(R)(x).

Another application of Proposition 3.3 is as follows: let F be a field of

characteristic 0. As we have seen in the homework, the function f : Z →
F defined by f (n) = n · 1 is a ring homomorphism. If char F = 0, the
homomorphism f is injective. Hence by Proposition 3.3 there is an induced
homomorphism f˜: Q → F . Its image is the set of all quotients in F of the
form n · 1/m · 1, with m 6= 0. In particular, the image of f˜ is a subfield of
F isomorphic to Q. Thus every field of characteristic 0 contains a subfield
isomorphic to Q, called the prime subfield. It is the smallest subfield of F ,
hence unique, and it can be described by starting with 1 and making sure
that we can perform the operations of addition and subtraction and then
automatically multiplication (to get the subring isomorphic to Z), and finally
division to get the subfield isomorphic to Q. Here “prime” has nothing to do
with prime numbers but simply means that the field Q is a basic, indivisible
object.
A similar statement holds if F is a field of positive characteristic, say
char F = p where p is a prime number. In this case, the function f : Z → F
defined by f (n) = n · 1 is still a ring homomorphism, but its kernel is hpi
and hence its image, as an abelian group, is isomorphic to Z/pZ. The fact
that f is a ring homomorphism implies that the image of f , as a ring, is

8
isomorphic to Z/pZ = Fp . Thus, every field of characteristic p contains
a subfield isomorphic to Fp , again called the prime subfield. The fields
Q and Fp are more generally called prime fields. They contain no proper
subfields, and every field F contains a unique subfield isomorphic either to
Q, if char F = 0, or to Fp , if char F = p.