Basics of Algebra, Topology, and Differential Calculus by Jean Gallier - HTML preview

PLEASE NOTE: This is an HTML preview only and some elements such as links or page numbers may be incorrect.
Download the book in PDF, ePub, Kindle for a complete version.

Chapter 20

Polynomials, Ideals and PID’s

20.1

Multisets

This chapter contains a review of polynomials and their basic properties. First, multisets

are defined. Polynomials in one variable are defined next. The notion of a polynomial

function in one argument is defined. Polynomials in several variable are defined, and so is

the notion of a polynomial function in several arguments. The Euclidean division algorithm is

presented, and the main consequences of its existence are derived. Ideals are defined, and the

characterization of greatest common divisors of polynomials in one variables (gcd’s) in terms

of ideals is shown. We also prove the Bezout identity. Next, we consider the factorization of

polynomials in one variables into irreducible factors. The unique factorization of polynomials

in one variable into irreducible factors is shown. Roots of polynomials and their multiplicity

are defined. It is shown that a nonnull polynomial in one variable and of degree m over an

integral domain has at most m roots. The chapter ends with a brief treatment of polynomial

interpolation: Lagrange, Newton, and Hermite interpolants are introduced.

In this chapter, it is assumed that all rings considered are commutative. Recall that a

(commutative) ring A is an integral domain (or an entire ring) if 1 = 0, and if ab = 0, then

either a = 0 or b = 0, for all a, b ∈ A. This second condition is equivalent to saying that if

a = 0 and b = 0, then ab = 0. Also, recall that a = 0 is not a zero divisor if ab = 0 whenever

b = 0. Observe that a field is an integral domain.

Our goal is to define polynomials in one or more indeterminates (or variables) X1, . . . , Xn,

with coefficients in a ring A. This can be done in several ways, and we choose a definition

that has the advantage of extending immediately from one to several variables. First, we

need to review the notion of a (finite) multiset.

Definition 20.1. Given a set I, a (finite) multiset over I is any function M : I → N such

that M (i) = 0 for finitely many i ∈ I. The multiset M such that M(i) = 0 for all i ∈ I is

the empty multiset, and it is denoted by 0. If M (i) = k = 0, we say that i is a member of

M of multiplicity k. The union M1 + M2 of two multisets M1 and M2 is defined such that

(M1 + M2)(i) = M1(i) + M2(i), for every i ∈ I. If I is finite, say I = {1, . . . , n}, the multiset

531

532

CHAPTER 20. POLYNOMIALS, IDEALS AND PID’S

M such that M (i) = ki for every i, 1 ≤ i ≤ n, is denoted by k1 · 1 + · · · + kn · n, or more

simply, by (k1, . . . , kn), and deg(k1 · 1 + · · · + kn · n) = k1 + · · · + kn is the size or degree of

M . The set of all multisets over I is denoted by

(I)

(n)

N

, and when I = {1, . . . , n}, by N .

Intuitively, the order of the elements of a multiset is irrelevant, but the multiplicity of

each element is relevant, contrary to sets. Every i ∈ I is identified with the multiset Mi such

that M

(1)

i(i) = 1 and Mi(j) = 0 for j = i. When I = {1}, the set N

of multisets k · 1 can be

identified with N and {1}∗. We will denote k · 1 simply by k.

However, beware that when n ≥ 2, the set (n)

N

of multisets cannot be identified with the

set of strings in {1, . . . , n}∗, because multiset union is commutative, but concatenation

of strings in {1, . . . , n}∗ is not commutative when n ≥ 2. This is because in a multiset

k1 · 1 + · · · + kn · n, the order is irrelevant, whereas in a string, the order is relevant. For

example, 2 · 1 + 3 · 2 = 3 · 2 + 2 · 1, but 11222 = 22211, as strings over {1, 2}.

Nevertherless,

(n)

n

N

and the set N of ordered n-tuples under component-wise addition

are isomorphic under the map

k1 · 1 + · · · + kn · n → (k1, . . . , kn).

Thus, since the notation (k1, . . . , kn) is less cumbersome that k1 · 1 + · · · + kn · n, it will be

preferred. We just have to remember that the order of the ki is really irrelevant.

But when I is infinite, beware that

(I)

I

N

and the set N of ordered I-tuples are not

isomorphic.

We are now ready to define polynomials.

20.2

Polynomials

We begin with polynomials in one variable.

Definition 20.2. Given a ring A, we define the set PA(1) of polynomials over A in one

variable as the set of functions P : N → A such that P (k) = 0 for finitely many k ∈ N. The

polynomial such that P (k) = 0 for all k ∈ N is the null (or zero) polynomial and it is denoted

by 0. We define addition of polynomials, multiplication by a scalar, and multiplication of

polynomials, as follows: Given any three polynomials P, Q, R ∈ PA(1), letting ak = P (k),

bk = Q(k), and ck = R(k), for every k ∈ N, we define R = P + Q such that

ck = ak + bk,

R = λP such that

ck = λak,

where λ ∈ A,

20.2. POLYNOMIALS

533

and R = P Q such that

ck =

aibj.

i+j=k

We define the polynomial ek such that ek(k) = 1 and ek(i) = 0 for i = k. We also denote

e0 by 1 when k = 0. Given a polynomial P , the ak = P (k) ∈ A are called the coefficients

of P . If P is not the null polynomial, there is a greatest n ≥ 0 such that an = 0 (and thus,

ak = 0 for all k > n) called the degree of P and denoted by deg(P ). Then, P is written

uniquely as

P = a0e0 + a1e1 + · · · + anen.

When P is the null polynomial, we let deg(P ) = −∞.

There is an injection of A into PA(1) given by the map a → a1 (recall that 1 denotes e0).

There is also an injection of N into PA(1) given by the map k → ek. Observe that ek = ek1

(with e01 = e0 = 1). In order to alleviate the notation, we often denote e1 by X, and we call

X a variable (or indeterminate). Then, ek = ek1 is denoted by Xk. Adopting this notation,

given a nonnull polynomial P of degree n, if P (k) = ak, P is denoted by

P = a0 + a1X + · · · + anXn,

or by

P = anXn + an−1Xn−1 + · · · + a0,

if this is more convenient (the order of the terms does not matter anyway). Sometimes, it

will also be convenient to write a polynomial as

P = a0Xn + a1Xn−1 + · · · + an.

The set PA(1) is also denoted by A[X] and a polynomial P may be denoted by P (X).

In denoting polynomials, we will use both upper-case and lower-case letters, usually, P, Q,

R, S, p, q, r, s, but also f, g, h, etc., if needed (as long as no ambiguities arise).

Given a nonnull polynomial P of degree n, the nonnull coefficient an is called the leading

coefficient of P . The coefficient a0 is called the constant term of P . A polynomial of the

form akXk is called a monomial. We say that akXk occurs in P if ak = 0. A nonzero

polynomial P of degree n is called a monic polynomial (or unitary polynomial, or monic) if

an = 1, where an is its leading coefficient, and such a polynomial can be written as

P = Xn + an−1Xn−1 + · · · + a0

or

P = Xn + a1Xn−1 + · · · + an.

The choice of the variable X to denote e1 is standard practice, but there is nothing special

about X. We could have chosen Y , Z, or any other symbol, as long as no ambiguities

arise.

534

CHAPTER 20. POLYNOMIALS, IDEALS AND PID’S

Formally, the definition of PA(1) has nothing to do with X. The reason for using X is

simply convenience. Indeed, it is more convenient to write a polynomial as P = a0 + a1X +

· · · + anXn rather than as P = a0e0 + a1e1 + · · · + anen.

We have the following simple but crucial proposition.

Proposition 20.1. Given two nonnull polynomials P (X) = a0 +a1X +· · ·+amXm of degree

m and Q(X) = b0 + b1X + · · · + bnXn of degree n, if either am or bn is not a zero divisor,

then ambn = 0, and thus, P Q = 0 and

deg(P Q) = deg(P ) + deg(Q).

In particular, if A is an integral domain, then A[X] is an integral domain.

Proof. Since the coefficient of Xm+n in P Q is ambn, and since we assumed that either am or

an is not a zero divisor, we have ambn = 0, and thus, P Q = 0 and

deg(P Q) = deg(P ) + deg(Q).

Then, it is obvious that A[X] is an integral domain.

It is easily verified that A[X] is a commutative ring, with multiplicative identity 1X0 = 1.

It is also easily verified that A[X] satisfies all the conditions of Definition 2.9, but A[X] is

not a vector space, since A is not necessarily a field.

A structure satisfying the axioms of Definition 2.9 when K is a ring (and not necessarily a

field) is called a module. As we mentioned in Section 4.2, we will not study modules because

they fail to have some of the nice properties that vector spaces have, and thus, they are

harder to study. For example, there are modules that do not have a basis.

However, when the ring A is a field, A[X] is a vector space. But even when A is just a

ring, the family of polynomials (Xk)k∈ is a basis of A[X], since every polynomial P (X) can

N

be written in a unique way as P (X) = a0 + a1X + · · · + anXn (with P (X) = 0 when P (X)

is the null polynomial). Thus, A[X] is a free module.

Next, we want to define the notion of evaluating a polynomial P (X) at some α ∈ A. For

this, we need a proposition.

Proposition 20.2. Let A, B be two rings and let h : A → B be a ring homomorphism.

For any β ∈ B, there is a unique ring homomorphism ϕ : A[X] → B extending h such that

ϕ(X) = β, as in the following diagram (where we denote by h+β the map h+β : A∪{X} → B

such that (h + β)(a) = h(a) for all a ∈ A and (h + β)(X) = β):

A ∪ {X} ι /

h+β

%▲

A[X]

ϕ

B

20.2. POLYNOMIALS

535

Proof. Let ϕ(0) = 0, and for every nonull polynomial P (X) = a0 + a1X + · · · + anXn, let

ϕ(P (X)) = h(a0) + h(a1)β + · · · + h(an)βn.

It is easily verified that ϕ is the unique homomorphism ϕ : A[X] → B extending h such that

ϕ(X) = β.

Taking A = B in Proposition 20.2 and h : A → A the identity, for every β ∈ A, there

is a unique homomorphism ϕβ : A[X] → A such that ϕβ(X) = β, and for every polynomial

P (X), we write ϕβ(P (X)) as P (β) and we call P (β) the value of P (X) at X = β. Thus, we

can define a function PA : A → A such that PA(β) = P (β), for all β ∈ A. This function is

called the polynomial function induced by P .

More generally, PB can be defined for any (commutative) ring B such that A ⊆ B. In

general, it is possible that PA = QA for distinct polynomials P, Q. We will see shortly

conditions for which the map P → PA is injective. In particular, this is true for A = R (in

general, any infinite integral domain). We now define polynomials in n variables.

Definition 20.3. Given n ≥ 1 and a ring A, the set PA(n) of polynomials over A in n

variables is the set of functions P :

(n)

N

→ A such that P (k1, . . . , kn) = 0 for finitely many

(k

(n)

1, . . . , kn) ∈ N

. The polynomial such that P (k1, . . . , kn) = 0 for all (k1, . . . , kn) is

the null (or zero) polynomial and it is denoted by 0. We define addition of polynomials,

multiplication by a scalar, and multiplication of polynomials, as follows: Given any three

polynomials P, Q, R ∈ PA(n), letting a(k1,...,kn) = P (k1, . . . , kn), b(k1,...,kn) = Q(k1, . . . , kn),

c

(n)

(k

, we define R = P + Q such that

1,...,kn) = R(k1, . . . , kn), for every (k1, . . . , kn) ∈ N

c(k1,...,kn) = a(k1,...,kn) + b(k1,...,kn),

R = λP , where λ ∈ A, such that

c(k1,...,kn) = λa(k1,...,kn),

and R = P Q, such that

c(k

a

1,...,kn) =

(i1,...,in)b(j1,...,jn).

(i1,...,in)+(j1,...,jn)=(k1,...,kn)

For every (k

(n)

1, . . . , kn) ∈ N

, we let e(k1,...,kn) be the polynomial such that

e(k1,...,kn)(k1, . . . , kn) = 1 and e(k1,...,kn)(h1, . . . , hn) = 0,

for (h1, . . . , hn) = (k1, . . . , kn). We also denote e(0,...,0) by 1. Given a polynomial P , the

a(k1,...,kn) = P (k1, . . . , kn) ∈ A, are called the coefficients of P . If P is not the null polynomial,

there is a greatest d ≥ 0 such that a

(n)

(k

, with d =

1,...,kn)

= 0 for some (k1, . . . , kn) ∈ N

k1 + · · · + kn, called the total degree of P and denoted by deg(P ). Then, P is written

uniquely as

P =

a(k1,...,kn)e(k1,...,kn).

(k

(n)

1,...,kn)∈N

When P is the null polynomial, we let deg(P ) = −∞.

536

CHAPTER 20. POLYNOMIALS, IDEALS AND PID’S

There is an injection of A into PA(n) given by the map a → a1 (where 1 denotes e(0,...,0)).

There is also an injection of (n)

N

into PA(n) given by the map (h1, . . . , hn) → e(h1,...,hn). Note

that e(h1,...,hn)e(k1,...,kn) = e(h1+k1,...,hn+kn). In order to alleviate the notation, let X1, . . . , Xn

be n distinct variables and denote e(0,...,0,1,0...,0), where 1 occurs in the position i, by Xi

(where 1 ≤ i ≤ n). With this convention, in view of e(h1,...,hn)e(k1,...,kn) = e(h1+k1,...,hn+kn), the

polynomial e(k1,...,kn) is denoted by Xk1

1 · · · X kn

n

(with e(0,...,0) = X01 · · · X0n = 1) and it is called

a primitive monomial . Then, P is also written as

P =

a(k1,...,kn)Xk1

1 · · · X kn

n .

(k

(n)

1,...,kn)∈N

We also denote PA(n) by A[X1, . . . , Xn]. A polynomial P ∈ A[X1, . . . , Xn] is also denoted

by P (X1, . . . , Xn).

As in the case n = 1, there is nothing special about the choice of X1, . . . , Xn as variables

(or indeterminates). It is just a convenience. After all, the construction of PA(n) has nothing

to do with X1, . . . , Xn.

Given a nonnull polynomial P of degree d, the nonnull coefficients a(k1,...,kn) = 0 such

that d = k1 + · · · + kn are called the leading coefficients of P . A polynomial of the form

a(k1,...,kn)Xk1

1 · · · X kn

n

is called a monomial . Note that deg(a(k1,...,kn)Xk1

1 · · · X kn

n ) = k1+· · ·+kn.

Given a polynomial

P =

a(k1,...,kn)Xk1

1 · · · X kn

n ,

(k

(n)

1,...,kn)∈N

a monomial a(k1,...,kn)Xk1

1 · · · X kn

n

occurs in the polynomial P if a(k1,...,kn) = 0.

A polynomial

P =

a(k1,...,kn)Xk1

1 · · · X kn

n

(k

(n)

1,...,kn)∈N

is homogeneous of degree d if

deg(Xk1

1 · · · X kn

n ) = d,

for every monomial a(k1,...,kn)Xk1

1 · · · X kn

n

occurring in P . If P is a polynomial of total degree

d, it is clear that P can be written uniquely as

P = P (0) + P (1) + · · · + P (d),

where P (i) is the sum of all monomials of degree i occurring in P , where 0 ≤ i ≤ d.

It is easily verified that A[X1, . . . , Xn] is a commutative ring, with multiplicative identity

1X01 · · · X0n = 1. It is also easily verified that A[X] is a module. When A is a field, A[X] is

a vector space.

Even when A is just a ring, the family of polynomials

(Xk1

1 · · · X kn

n )(k

(n)

1,...,kn)∈N

20.2. POLYNOMIALS

537

is a basis of A[X1, . . . , Xn], since every polynomial P (X1, . . . , Xn) can be written in a unique

way as

P (X1, . . . , Xn) =

a(k1,...,kn)Xk1

1 · · · X kn

n .

(k

(n)

1,...,kn)∈N

Thus, A[X1, . . . , Xn] is a free module.

Remark: The construction of Definition 20.3 can be immediately extended to an arbitrary

set I, and not just I = {1, . . . , n}. It can also be applied to monoids more general that (I)

N

.

Proposition 20.2 is generalized as follows.

Proposition 20.3. Let A, B be two rings and let h : A → B be a ring homomorphism. For

any β = (β1, . . . , βn) ∈ Bn, there is a unique ring homomorphism ϕ : A[X1, . . . , Xn] → B

extending h such that ϕ(Xi) = βi, 1 ≤ i ≤ n, as in the following diagram (where we denote

by h + β the map h + β : A ∪ {X1, . . . , Xn} → B such that (h + β)(a) = h(a) for all a ∈ A

and (h + β)(Xi) = βi, 1 ≤ i ≤ n):

A ∪ {X1, . . . , Xn} ι /

h+β

)❚

A[X1, . . . , Xn]

ϕ

B

Proof. Let ϕ(0) = 0, and for every nonull polynomial

P (X1, . . . , Xn) =

a(k1,...,kn)Xk1

1 · · · X kn

n ,

(k

(n)

1,...,kn)∈N

let

ϕ(P (X1, . . . , Xn)) =

h(a(k1,...,kn))βk1

1 · · · βkn

n .

It is easily verified that ϕ is the unique homomorphism ϕ : A[X1, . . . , Xn] → B extending h

such that ϕ(Xi) = βi.

Taking A = B in Proposition 20.3 and h : A → A the identity, for every β1, . . . , βn ∈ A,

there is a unique homomorphism ϕ : A[X1, . . . , Xn] → A such that ϕ(Xi) = βi, and for

every polynomial P (X1, . . . , Xn), we write ϕ(P (X1, . . . , Xn)) as P (β1, . . . , βn) and we call

P (β1, . . . , βn) the value of P (X1, . . . , Xn) at X1 = β1, . . . , Xn = βn. Thus, we can define a

function PA : An → A such that PA(β1, . . . , βn) = P (β1, . . . , βn), for all β1, . . . , βn ∈ A. This

function is called the polynomial function induced by P .

More generally, PB can be defined for any (commutative) ring B such that A ⊆ B. As

in the case of a single variable, it is possible that PA = QA for distinct polynomials P, Q.

We will see shortly that the map P → PA is injective when A = R (in general, any infinite

integral domain).

538

CHAPTER 20. POLYNOMIALS, IDEALS AND PID’S

Given any nonnull polynomial P (X1, . . . , Xn) =

(k

(n) a(k

1,...,kn)∈N

1,...,kn)X k1

1 · · · X kn

n

in

A[X1, . . . , Xn], where n ≥ 2, P (X1, . . . , Xn) can be uniquely written as

P (X1, . . . , Xn) =

Qk (X

n

1, . . . , Xn−1)X kn

n ,

where each polynomial Qk (X

n

1, . . . , Xn−1) is in A[X1, . . . , Xn−1]. Thus, even if A is a field,

A[X1, . . . , Xn−1] is not a field, which confirms that it is useful (and necessary!) to consider

polynomials over rings that are not necessarily fields.

It is not difficult to show that A[X1, . . . , Xn] and A[X1, . . . , Xn−1][Xn] are isomorphic

rings. This way, it is often possible to prove properties of polynomials in several variables

X1, . . . , Xn, by induction on the number n of variables. For example, given two nonnull

polynomials P (X1, . . . , Xn) of total degree p and Q(X1, . . . , Xn) of total degree q, since we

assumed that A is an integral domain, we can prove that

deg(P Q) = deg(P ) + deg(Q),

and that A[X1, . . . , Xn] is an integral domain.

Next, we will consider the division of polynomials (in one variable).

20.3

Euclidean Division of Polynomials

We know that every natural number n ≥ 2 can be written uniquely as a product of powers of

prime numbers and that prime numbers play a very important role in arithmetic. It would be

nice if every polynomial could be expressed (uniquely) as a product of “irreducible” factors.

This is indeed the case for polynomials over a field. The fact that there is a division algorithm

for the natural numbers is essential for obtaining many of the arithmetical properties of the

natural numbers. As we shall see next, there is also a division algorithm for polynomials in

A[X], when A is a field.

Proposition 20.4. Let A be a ring, let f (X), g(X) ∈ A[X] be two polynomials of degree

m = deg(f ) and n = deg(g) with f (X) = 0, and assume that the leading coefficient am of

f (X) is invertible. Then, there exist unique polynomials q(X) and r(X) in A[X] such that

g = f q + r

and

deg(r) < deg(f ) = m.

Proof. We first prove the existence of q and r. Let

f = amXm + am−1Xm−1 + · · · + a0,

and

g = bnXn + bn−1Xn−1 + · · · + b0.

If n < m, then let q = 0 and r = g. Since deg(g) < deg(f ) and r = g, we have deg(r) <

deg(f ).

20.3. EUCLIDEAN DIVISION OF POLYNOMIALS

539

If n ≥ m, we proceed by induction on n. If n = 0, then g = b0, m = 0, f = a0 = 0, and

we let q = a−1

0 b0 and r = 0. Since deg(r) = deg(0) = −∞ and deg(f ) = deg(a0) = 0 because

a0 = 0, we have deg(r) < deg(f ).

If n ≥ 1, since n ≥ m, note that

g1(X) = g(X) − bna−1

m X n−mf (X )

= bnXn + bn−1Xn−1 + · · · + b0 − bna−1

m X n−m(amX m + am−1X m−1 + · · · + a0)

is a polynomial of degree deg(g1) < n, since the terms bnXn and bna−1

m X n−mamX m of degree

n cancel out. Now, since deg(g1) < n, by the induction hypothesis, we can find q1 and r

such that

g1 = f q1 + r and deg(r) < deg(f ) = m,

and thus,

g1(X) = g(X) − bna−1

m X n−mf (X ) = f (X )q1(X ) + r(X ),

from which, letting q(X) = bna−1

m X n−m + q1(X ), we get

g = f q + r

and deg(r) < m = deg(f ).

We now prove uniqueness. If

g = f q1 + r1 = f q2 + r2,

with deg(r1) < deg(f ) and deg(r2) < deg(f ), we get

f (q1 − q2) = r2 − r1.

If q2 − q1 = 0, since the leading coefficient am of f is invertible, by Proposition 20.1, we have

deg(r2 − r1) = deg(f(q1 − q2)) = deg(f) + deg(q2 − q1),

and so, deg(r2−r1) ≥ deg(f), which contradicts the fact that deg(r1) < deg(f) and deg(r2) <

deg(f ). Thus, q1 = q2, and then also r1 = r2.

It should be noted that the proof of Proposition 20.4 actually provides an algorithm for

finding the quotient q and the remainder r of the division of g by f . This algorithm is

called the Euclidean algorithm, or division algorithm. Note that the division of g by f is

always possible when f is a monic polynomial, since 1 is invertible. Also, when A is a field,

am = 0 is always invertible, and thus, the division can always be performed. We say that f

divides g when r = 0 in the result of the division g = f q + r. We now draw some important

consequences of the existence of the Euclidean algorithm.

540

CHAPTER 20. POLYNOMIALS, IDEALS AND PID’S

20.4

Ideals, PID’s, and Greatest Common Divisors

First, we introduce the fundamental concept of an ideal.

Definition 20.4. Given a ring A, an ideal of A is any nonempty subset I of A satisfying

the following two properties:

(ID1) If a, b ∈ I, then b − a ∈ I.

(ID2) If a ∈ I, then ax ∈ I for every x ∈ A.

An ideal I is a principal ideal if there is some a ∈ I, called a generator, such that

I = {ax | x ∈ A}.

The equality I = {ax | x ∈ A} is also written as I = aA or as I =