4

Divisibility in integral domains

We have until this moment used Z and K[X] both as rich sources of examples, and also as prototypes: we have tried to develop ring theory in a way that captures the essential properties of these structures. In this section we will try to generalize the divisibility and factorization properties of these rings in a more general settings. For example we would like to generalize the Fundamental Theorem of Arithmetic and the Euclidean Algorithm from the case of Z to more general rings. We begin with introducing the appropriate terminology in the abstract setting of integral domains.

4.1

Divisibility

Denition 1. Let R be a ring, and let a, b R. We say that b divides a (and write b|a) if for some c R we have that a = bc. We will also call b a divisor of a in these circumstances. So for example the units are the divisors of 1, and everything divides 0. Denition 2. Let R be a ring, and let a R. An element a of R is called an associate of a if for some unit u R we have that a = ua. We remark that this is an equivalence relation. For example the associates of n Z are n. Note that unique factorization in Z is not really unique, for example 6 = 2(3) = (2)3. Of course this non uniqueness up to a sign is quite harmless. In the unique factorization theorem that we will prove later uniqueness will fail up to units rather than signs, which is still not bad. As we want to think of 2 and 2 as the same factor of 6 associate elements are the same as far as divisibility goes for general rings. Remark 4.1.1. If R is an integral domain and a, b R are such that a|b and b|a then a, b are associates. Indeed if both a, b are 0 they are clearly associates. Otherwise lets say that b = 0. Since a|b and b|a we have b = au and a = bv for some u, v R. Substituting a in the rst equality we get b = (bv)u b b(vu) = 0 b(1 vu) = 0 1 vu = 0 So u, v are units and a, b associates. Denition 3. Let R be a ring, and let a R, a = 0. We say that a is irreducible in R if a R and / a = xy = x is a unit, or y is a unit. Denition 4. Let R be a ring, and let a R \ {0} but a R . We say that a is prime in R if a|xy = a|x or a|y. Essentially, irreducible elements cant be factorised further; prime elements are those which only divide products of which they already divide one factor. In general we have this: Proposition 4.1.2 (Prime Irreducible). Let R be an integral domain, and let x R be prime. Then x is irreducible. 16

Proof. Suppose that x = yz. Then x = x1 so that x|x = yz. By the denition of prime we will get x|y or x|z; suppose the former, that y = xt say. Then x = yz = xtz, and so x(1tz) = 0. As x = 0 we get tz = 1, and z is a unit. Hence x is by denition irreducible. Denition 5. Let R be a ring, and let a, b R. We say that d is a highest common factor of a, b if (i) d|a and d|b; (ii) e|a and e|b implies that e|d. So for example in Z a highest common factor of 9 and 6 is (3); another is 3. Note that in general there is no guarantee that highest common factors exist. Proposition 4.1.3. Let R be an integral domain and let a, b R \ {0}. Suppose that d1 and d2 are highest common factors of a, b. Then d1 and d2 are associates. Proof. It is clear that neither d1 nor d2 is zero, as a and b are multiples of each. By condition (i) applied to d1 we see that d1 |a and d1 |b. So apply condition (ii) to d2 using e = d1 to get that d1 |d2 . Similarly we get d2 |d1 . So by remark 4.1.1 we have that d1 , d2 are associates. Proposition 4.1.4. Let R be an integral domain and let a, b R \ {0}. Suppose that d1 is a highest common factor of a, b and that d2 is an associate of d1 . Then d2 is a highest common factor too. Proof. Suppose that d2 = d1 u for some u such that uv = 1. For condition (i) note that a = d1 x1 and b = d1 y1 , so that a = d2 (vx1 ) and b = d2 (vy1 ). For condition (ii) suppose that e|a and e|b. Then we have that e|d1 , or d1 = ez. Then d2 = d1 u = ezu and we are done. In some rings there is a sensible way to choose a particular highest common factorin Z we usually choose the non-negative associate, in K[X] the monic associateand call it the highest common factor. But often we can work quite comfortably with the uncertainty of up to a unit multiple.

4.2

Euclidean Rings

In order to generalize the theorems of Arithmetic to rings we have to restrict to a special class of rings: Euclidean rings.

4.3

Denition

What we must do is express abstractly what is going on in the Division Algorithm. We divide a by b = 0 and get a quotient q, leaving a remainder r which is smaller in some way (size, degree, . . . ) than the divisor b. Denition 6. We say that R is a Euclidean Ring with Euclidean function d if (a) R is an integral domain; (b) the function d : R \ {0} N \ {0} satises: 17

(i) for all x, y R \ {0} we have d(xy)

d(y);

(ii) for all a, b R with b = 0 there exist q, r R such that a = bq + r and r = 0 or d(r) < d(b).

The element q = q(a, b) is called the quotient of a by b, and the element r = r(a, b) is called the remainder. In our customary rather slovenly way we will say R is a Euclidean ring when the function d is so obvious as to be understood without mention. However, see the last example below for a warning.

4.4
4.4.1

Examples
The Integers

The ring of integers, equipped with the function n |n| on the non-zero elements, is a Euclidean Ring: we have known this since we learned about division, and we proved that it is true in the rst year course1 . 4.4.2 Rational Polynomials

The ring Q[X], equipped with the degree function f deg f , is a Euclidean Ring: weve known this since we learned about long division, and we proved that it is true in the rst year course2 . Of course the same thing is true for polynomials over any eld. 4.4.3 The Gaussian Integers

Let Z[i] := {a + bi | a, b Z}. We call these the Gaussian integers. This subset of C is clearly a subring of C and as C has no zero-divisors it is actually an integral domain. How can we measure size, and nd a Euclidean function d? The obvious choice is d(a + bi) := |a + bi|2 , but does it satisfy the requirements? Condition (i) is easy: d() = ||2 = ||2 ||2 = ||2 d() and as || = a2 + b2 for integral a, b we get ||2 1. Condition (ii) is more complicated; the argument is important as it can be used for certain other Z[ n] and not just Z[ 1]. So let := a + bi and := c + di = 0 be in the ring. Then in Q[i] := {x + yi | x, y Q} we can rationalise the denominator and get that a + bi ac + bd bc ad = = 2 + 2 i =: x + yi c + di c + d2 c + d2 say.
1 2

On the course website there will be a proof. On the course website there will be a proof.

18

This is the exact quotient in Q[i], but what is the best we can do in Z[i]? The nearest we can get is the number := m + ni, where m is the integer nearest x, and n the integer nearest y; note that |x m| 1 and |y n| 1 . 2 2 Now = x + yi = m + ni + (x m) + (y n)i and multiplying by we get = (m + ni) + ((x m) + (y n)i) . Put as quotient q(, ) := m + ni Z[i]. As remainder we then would have r(, ) := ((x m) + (y n)i) = q(, ) Z[i]. We now compute d(((x m) + (y n)i) ) = | ((x m) + (y n)i) |2 = | ((x m) + (y n)i) |2 ||2 = ((x m)2 + (y n)2 ) ||2
1 2

< d() and see that condition (ii) is satised. 4.4.4 Fields

||2

Let K be a eld, and dene d : K \ {0} N \ {0} by d(x) = 1. It is then trivial to see that we have a (very dull) Euclidean Ring where all the remainders are 0. 4.4.5 The Integers, but not as we know them

The ring of integers, equipped with the function d(n) = the number of digits when |n| is expressed in base 2, is a Euclidean Ring. For condition (i), note that 2M + lower powers of 2 2N + lower powers of 2 = 2M +N + lower powers of 2 and so d(xy) = d(x) + d(y) d(y). For condition (ii), all is clear if a = 0, or indeed if d(a) < d(b); just take q = 0 and r = a. So argue by induction on d(a). Suppose that a = 2M + lower powers of 2 , and b = 2N + lower powers of 2 with = 1 and = 1. We are assuming M N , so M N b + 2M N b. The number a 2M N b requires fewer than consider a = a 2 M = d(a) binary digits, so we can nd q and r such that a 2M N b = b + r q and r = 0 or d(r) < d(b). Taking q = q + 2M N (= 2M N ) gives what we need. q This example shows that we need to take care when we say R is a Euclidean Ring. 19

4.5

Units and Associates

If we are interested in factorisations the rst thing we need to deal with are the factorisations of the identity: we must nd the units of the ring. So let R equipped with d be a Euclidean ring. Lemma 4.5.1. For all a R, d(a) d(1).

Proof. Condition (i) applied to a = a 1 gives this at once. Lemma 4.5.2. For all units u R , d(u) d(1). Proof. Condition (i) applied to 1 = v u gives this at once. Lemma 4.5.3. For all x R such that d(x) = d(1), we have that x is a unit. Proof. Use condition (ii) to get q and r such that 1 = xq + r with r = 0, or d(r) < d(x).

If r = 0 then we have by hypothesis d(r) < d(x) = d(1); this contradicts Lemma 1. So we get exact division, 1 = xq and x is a unit. To summarise: Proposition 4.5.4 (Units of a Euclidean ring). Let R equipped with d be a Euclidean ring; then the group of units is given by R = {x R | d(x) = d(1)} . In fact we may arrange things so that we have d(1) = 1. For suppose that d(1) = k + 1 for k N. Weve just seen that for all a R we have that d(a) d(1), so the function d : R \ {0} N \ {0} by d : a d(a) k is well-dened. It is clear that it also satises condition (i) and condition (ii) with the same quotient and remainder. The following is also useful about a Euclidean ring R: Lemma 4.5.5. Let u R and a R. Then d(ua) = d(a). Proof. Let v a be such that vu = 1. We then have by condition (i) that d(a) = d(vua) d(ua) d(a); equalities rule, and d(a) = d(ua).

20

The following subsections, establishing that the integers and polynomial rings over elds are Euclidean are included purely for completeness; they were covered in the rst year course.

Revision: integers
Let d : Z \ {0} N \ {0} be given by d(n) = |n|. For condition (i) we can use the properties of see that d(xy) = |xy| = |x||y| = |x|d(y) d(y). For condition (ii) note that and |n| we developed in Analysis I and

a = bq + r (a) = b(q) + (r) and |r| = | r|; so it is enough to deal with the case a 0. We can argue by (strong) induction on a; the result is true for a = 0 if we take q = 0 and r = 0. Indeed the result is true for a < |b|, just take q = 0 and r = a. For a |b| note that a = (a |b|) + |b|, so we can use the inductive hypothesis to get (a |b|) = b + r with r = 0 q or |r| < |b|. Taking q = q 1 as case may be completes the proof.

Revision: polynomials
Let d : K[X] \ {0} N \ {0} be given by d(f ) = deg(f ), the degree of f . For condition (i) note that if f = m ak X k and g = n bk X k , with am = 0 and k=0 k=0 m+n bn = 0 then f g = k=0 ck X k where cm+n = am bn . Hence deg f g = deg f + deg g deg g. For condition (ii), again it is clearly true if f = 0; just take q = r = 0. Otherwise note that if the top coecient of f is = 0 and the top coecient of g is = 0, then
1 f = gq + r f = 1 g q

1 r

1 1 1 clearly deg f = deg f , deg g = deg g and deg r = deg r. Hence we may assume that f, g are monic, that is have top coecients 1. Now we may argue by induction on deg f ; if deg f = 0 or more generally deg f < deg g we put q = 0 and r = f and are done. Otherwise, note that f (X) = f (X) X deg f deg g g(x) + X deg f deg g g(x), and that deg f (X) X deg f deg g g(x) < deg f . By the inductive hypothesis we can then nd q and r such that

f (X) X deg f deg g g(X) = g(X)(X) + r(X) q with deg r < deg g or r = 0. Now put q(X) = q (X) + X deg f deg g and we are done.

21

Which of the following are true? 1. If K is a eld then any non zero element of K divides any other non zero element of K. 2. If a, b are associates then a|b and b|a. 3. If a is irreducible and a, b are associates then b is also irreducible. 4. If a is prime and a, b are associates then b is also prime. 5. a|b if and only if (a) (b). 6. The ideal < X > + < Y > of R[X, Y ] is principal. 7. R[X, Y ] is a Euclidean domain. 8. If R is Euclidean ring with Euclidean function d then d(a) > d(1) for all a = 0, a R.

22