Rings and Ideals
Rings and Ideals
Rings and Ideals
In this chapter we introduce some abstract algebra in order to shed some light
on several ad-hoc constructions that we have employed previously.
In general, a ring is a set on which two compositions called addition and
multiplication are defined in such a way that certain axioms hold. In particular,
R should be a group with respect to addition and a monoid with respect to
multiplication; moreover, distributivity a(b + c) = ab + bc should hold.
Here, all our rings will be commutative (ab = ba) domains (ab = 0 implies
a = 0 or b = 0; the ring Z/6Z is not a domain because [2][3] = [0]) and will
have a multiplicative unit 1 (the ring 2Z of even numbers does not have a unit;
sometimes such objects are called rngs).
81
Proof. Let us first show that the norm function is multiplicative. This means
that N [(a + bi)(c + di)] = N (a + bi) · N (c + di), and is easily checked by com-
putation.
Now assume that we are given elements a = r + si and b + t + ui in Z[i];
then we need to find q, r ∈ Z[i] with a = bq + r and N (r) < N (b). Since N is
multiplicative, this is equivalent to the statement that for every p = ab ∈ Q(i) =
{x + yi : x, y ∈ Q} there is an element q ∈ Z[i] with N (p − q) < 1.
Now write p − q = x + yi for x, y ∈ Q, and let q = c + di with c, d ∈ Z and
|x − c| ≤ 21 , |y − d| ≤ 12 . Then N (p − q) ≤ 14 + 14 = 12 < 1.
√
This result can be generalized
√ somewhat: the rings Z[ m ] are Euclidean
with respect to ν(a + b m ) = |a2 − mb2 | for m = −2, 2, 3. In fact, there
are more values of m for which these rings are Euclidean, but the proofs soon
become very technical.
Proposition 9.3. Let K be a field. Then the ring K[X] of polynomials in one
variable X with coefficients from K is Euclidean with respect to ν(f ) = 2deg f .
Proof. Assume that a, b ∈ R are nonzero polynomials. Then we have to find
q, r ∈ R with a = bq + r and deg r < deg b. We do this by induction and long
division.
First observe that the claim is trivial if deg a < deg b; thus we may assume
that deg a ≥ deg b. Then the claim is trivial if deg a = 0, since this implies
deg b = 0, hence b is a nonzero constant, hence a unit, and we can write a = bq+0
with q = ab−1 .
Now assume that the claim is true for all polynomials a with deg a < m, and
write a = am xm + am−1 xm−1 + . . . + a0 and b = an xn + bn−1 xn−1 + . . . + b0
with m ≥ n. Then a and b · q1 with q1 = abm n
X m−n are polynomials of degree m
with the same leading coefficient am , hence r1 = a − bq1 is a polynomial with
degree deg r1 < m. By induction assumption, there exist polynomials q, r with
r1 = bq + r and deg r < deg b. But now a = bq1 + r1 = b(q1 + q) + r, and this
proves the theorem.
9.2 Ideals
Our goal is to show that Euclidean rings are UFDs. This will have concrete
applications; apart from showing again that e.g. Z has unique factorization, the
fact that Z[i] is a UFD implies that every prime p ≡ 1 mod 4 is the sum of two
integral squares. Even Lagrange’s 4-squares theorem (every positive integer is
the sum of 4 integral squares) can be proved by showing that the division algebra
of quaternions A = Q(i, j, k) with i2 = j 2 = k 2 = −1 and ij = −ji, ij = k
contains a (left) Euclidean subring.
The proof that Euclidean rings are UFDs becomes simpler upon introducing
another type of rings: principal ideal rings (PIDs). Thus what we actually will
prove are the inclusions
82
Let R be a ring. A subring I of R is called an ideal if IR ⊆ I, i.e., if ir ∈ I
for all i ∈ I and all r ∈ R.
Let me give you a few examples:
• In any ring, the set (a) = {ar : r ∈ R} is an ideal for any a ∈ R. Such
ideals are called principal ideals. In particular, R = (1) and (0) are ideals.
83
does not converge
√ in Q (for example, take a sequence of rational numbers
“converging” to 2); then (an )(bn ) does not converge.
Finally, let N denote the set of null sequences, i.e., sequences converging
to 0. They form a subring of L, and actually form an ideal in L, in C and
even in B.
We have seen examples of this before when we showed that (3, 5) = (1) and
(6, 9) = (3).
84
9.3 Principal Ideal Domains
Now we claim
Theorem 9.9. Every Euclidean domain is a PID.
Proof. Let I be an ideal in the Euclidean ring R. If I = (0) we are done; thus
assume that I is not the zero ideal. Let a ∈ I be a nonzero element with minimal
f (a), where f is the Euclidean function. We claim that I = (a).
In fact, let b ∈ I and write b = aq + r with f (r) < f (a); since a ∈ I and I is
an ideal we know that aq ∈ I, hence r = b − aq ∈ I. By the definition of a we
must have r = 0, and this shows that every element of I is a multiple of a, i.e.,
I = (a).
This provides us with many (but not all) PIDs. In our proof of unique
factorization in Z, the main problem was showing that irreducibles are prime.
In PIDs, we get this for free:
Proposition 9.10. In any PID irreducible elements are prime.
Proof. Let p ∈ R be irreducible, and assume that p | ab. If p | a we are done, so
assume that p - a. We claim that (a, p) = (1) = R. In fact, write (d) = (a, p).
Then d | p, hence p = dr for d, r ∈ R. Since p is irreducible, d or r must be a
unit. If d is a unit, then (a, p) = (1) as claimed, and if r is a unit, then (d) = (p),
hence (a, p) = (p) and finally p | a: contradiction.
Thus (a, p) = (1), hence there exist r, s ∈ R with ar + ps = 1. But then
b = abr + aps, and since p | ab, p divides the right hand side and therefore b.
Next we have to show that every nonzero nonunit in a PID has a factorization
into
√ irreducibles.
√ √ This is not at all obvious: consider e.g. the domain D =
n
Z[ 2, 4 2, 8 2, . . .] containing Z and all roots
√ √ 21/2 for√n ≥ 1.
√ Then
√ 2 is not a
4 4
unit,
√ and it is not irreducible because 2 = 2· 2. But 2 = 2· 2 shows that
2 is also reducible, and this process can be continued indefinitely: although
2 is a nonunit, it is not a product of irreducibles because none of its factors is
irreducible. In PIDs, this does not happen:
Proposition 9.11. Let R be a PID. Then every a ∈ R \ {0} has a factorization
into a unit times irreducible elements.
Proof. If a is a unit, we are done. If a is a nonunit then we claim that a has an
irreducible factor. This is clear if a is irreducible; if not then it has a nontrivial
factorization a = a1 b1 . If a1 is irreducible, we are done; if not, then there is a
nontrivial factorization a1 = a2 b2 etc. In this way we get a sequence of elements
a1 , a2 , . . . with . . . , a3 | a2 , a2 | a1 , a1 | a. Consider the ideal I = (a, a1 , a2 , . . .).
Since R is a PID, there is a c ∈ R with I = (c). Since I is the union of the
ideals (a), (a1 ), (a2 ), . . . , c must be an element of one of these, say c ∈ (am ).
But then (c) ⊆ (am ) and (am ) ⊆ I = (c) imply that I = (am ). Now am+1 | am ,
as well as am | am+1 because am+1 ∈ I = (am ): this implies that am and am+1
differ by a unit, hence am = am+1 bm+1 is not a nontrivial factorization.
85
Thus we have shown that every nonzero nonunit a is divisible by an irre-
ducible element. We now claim that a has a factorization into irreducibles. In
fact, write a = a1 b1 with a1 irreducible. If b1 is irreducible, we are done; if
not, write b1 = a2 b2 with a2 irreducible and continue. By the same argument
as above this process must terminate, and after finitely many steps we have a
factorization of a into irreducibles.
Now we are ready to prove
Theorem 9.12. Every PID is a UFD.
Proof. We have already shown the following two facts:
We define the norm of an ideal I by N (I) = #R/I. Note that the norm of an
ideal might be infinite; for example, Z/(0) has infinitely many elements (distinct
integers determine distinct residue classes modulo (0); similarly, Z[X]/(X) is
infinite). Since R/I is a ring, we can form its unit group (R/I)× . We now
define Euler’s phi function for ideals in R by Φ(I) = #(R/I)× .
In the case R = Z we have proved that φ(pn ) = (p − 1)pn−1 for positive
primes (or φ(p) = (|p| − 1)|p|n−1 for arbitrary primes). We did this by counting
all the elements in Z/pn Z (there were pn of them) and subtracting the number
classes represented by multiples of p (there are pn−1 of them).
86
After this excursion into the depths of abstract algebra we now return to
number theory: in the next two chapters we will study the arithmetic of two
UFDs, namely the ring Z[i] of Gaussian integers and the ring Fp [X] of polyno-
mials with coefficients in the finite field Fp = Z/pZ.
Exercises
9.1 Let R be the ring of continuous functions R −→ R, where addition and multipli-
cation are defined pointwise.
87