2.3 Multiple roots

Even when polynomials in F[X]F[X] have no common factor in F[X]F[X], one might expect that they could acquire a common factor in Ω[X]\Omega[X] for some ΩF\Omega\supset F. In fact, this doesn’t happen — greatest common divisors don’t change when the field is extended.

Proposition 2.10

Let ff and gg be polynomials in F[X]F[X], and let Ω\Omega be an extension of FF. If r(X)r(X) is the gcd of ff and gg computed in F[X]F[X], then it is also the gcd of ff and gg in Ω[X]\Omega[X]. In particular, distinct monic irreducible polynomials in F[X]F[X] do not acquire a common root in any extension of F.F.

Proof.

Let rF(X)r_{F}(X) and rΩ(X)r_{\Omega}(X) be the greatest common divisors of ff and gg in F[X]F[X] and Ω[X]\Omega[X] respectively. Certainly rF(X)|rΩ(X)r_{F}(X)|r_{\Omega}(X) in Ω[X]\Omega[X], but Euclid’s algorithm (1.8) shows that there are polynomials aa and bb in F[X]F[X] such that

a(X)f(X)+b(X)g(X)=rF(X),a(X)f(X)+b(X)g(X)=r_{F}(X),

and so rΩ(X)r_{\Omega}(X) divides rF(X)r_{F}(X) in Ω[X]\Omega[X].

For the second statement, note that the hypotheses imply that gcd(f,g)=1\gcd(f,g)=1 (in F[X]F[X]), and so ff and gg can’t acquire a common factor in any extension field.

The proposition allows us to speak of the greatest common divisor of ff and gg without reference to a field.

Let fF[X]f\in F[X]. Then ff splits into linear factors

equation (4) (4)
f(X)=ai=1r(Xαi)mi, αi distinct, mi1i=1rmi=deg(f),f(X)=a\prod_{i=1}^{r}(X-\alpha_{i})^{m_{i}},\text{ }\alpha_{i}\text{ distinct, }m_{i}\geq 1\text{, }\sum_{i=1}^{r}m_{i}=\deg(f),

in E[X]E[X] for some extension EE of FF (see 2.4). We say that αi\alpha_{i} is a root of ff of multiplicity mim_{i} in EE. If mi>1m_{i}>1, then αi\alpha_{i} is said to be a multiple root of ff, and otherwise it is a simple root.

I claim that the unordered sequence of integers m1,,mrm_{1},\ldots,m_{r} in (4) is independent of the extension EE chosen to split ff. Certainly, it is unchanged when EE is replaced with its subfield F[α1,,αr]F[\alpha_{1},\ldots,\alpha_{r}], and so we may suppose that EE is a splitting field for ff. Let EE and EE^{\prime} be splitting fields for FF, and suppose that f(X)=ai=1r(Xαi)mif(X)=a\prod_{i=1}^{r}(X-\alpha_{i})^{m_{i}} in E[X]E[X] and f(X)=ai=1r(Xαi)mif(X)=a^{\prime}\prod_{i=1}^{r^{\prime}}(X-\alpha_{i}^{\prime})^{m_{i}^{\prime}} in E[X]E^{\prime}[X]. Let φ:EE\varphi\colon E\rightarrow E^{\prime} be an FF-isomorphism, which exists by (2.7b), and extend it to an isomorphism E[X]E[X]E[X]\rightarrow E^{\prime}[X] by sending XX to XX. Then φ\varphi maps the factorization of ff in E[X]E[X] onto a factorization

f(X)=φ(a)i=1r(Xφ(αi))mif(X)=\varphi(a)\prod_{i=1}^{r}(X-\varphi(\alpha_{i}))^{m_{i}}

in E[X]E^{\prime}[X]. By unique factorization, this coincides with the earlier factorization in E[X]E^{\prime}[X] up to a renumbering of the αi\alpha_{i}. Therefore r=rr=r^{\prime}, and

{m1,,mr}={m1,,mr}.\{m_{1},\ldots,m_{r}\}=\{m_{1}^{\prime},\ldots,m_{r}^{\prime}\}.

We say that ff has a multiple root when at least one of the mi>1m_{i}>1, and that ff has only simple roots when all mi=1m_{i}=1. Thus “ff has a multiple root” means “ff has a multiple root in one, hence every, extension of FF splitting ff”, and similarly for ``f``f has only simple roots”.

We wish to determine when a polynomial has a multiple root. If ff has a multiple factor in F[X]F[X], say f=fi(X)mif=\prod f_{i}(X)^{m_{i}} with some mi>1m_{i}>1, then obviously it will have a multiple root. If f=fif=\prod f_{i} with the fif_{i} distinct monic irreducible polynomials, then Proposition 2.10 shows that ff has a multiple root if and only if at least one of the fif_{i} has a multiple root. Thus, it suffices to determine when an irreducible polynomial has a multiple root.

Example 2.11

Let FF be of characteristic p0p\neq 0, and assume that FF contains an element aa that is not a ppth-power, for example, a=Ta=T in the field 𝔽(T)p.\mathbb{F}{}_{p}(T). Then XpaX^{p}-a is irreducible in F[X]F[X], but by 1.4 we have Xpa=(Xα)pX^{p}-a=(X-\alpha)^{p} in its splitting field. Thus an irreducible polynomial can have multiple roots.

The derivative of a polynomial f(X)=aiXif(X)=\sum a_{i}X^{i} is defined to be f(X)=iaiXi1f^{\prime}(X)=\sum ia_{i}X^{i-1}. The usual rules for differentiating sums and products still hold, but note that in characteristic pp the derivative of XpX^{p} is zero.

Proposition 2.12

For a nonconstant irreducible polynomial ff in F[X]F[X], the following statements are equivalent:

  1. (a)

    ff has a multiple root;

  2. (b)

    gcd(f,f)1\gcd(f,f^{\prime})\neq 1;

  3. (c)

    FF has nonzero characteristic pp and ff is a polynomial in XpX^{p};

  4. (d)

    all the roots of ff are multiple.

Proof.

(a) \Rightarrow (b). Let α\alpha be a multiple root of ff, and write f=(Xα)mg(X)f=(X-\alpha)^{m}g(X), m>1m>1, in some extension field. Then

equation (5) (5)
f(X)=m(Xα)m1g(X)+(Xα)mg(X).f^{\prime}(X)=m(X-\alpha)^{m-1}g(X)+(X-\alpha)^{m}g^{\prime}(X).

Hence ff and ff^{\prime} have XαX-\alpha as a common factor.

(b) \Rightarrow (c). As ff is irreducible and deg(f)<deg(f)\deg(f^{\prime})<\deg(f),

gcd(f,f)1f=0.\gcd(f,f^{\prime})\neq 1\implies f^{\prime}=0.

Let f=a0++adXdf=a_{0}+\cdots+a_{d}X^{d}, d1d\geq 1. Then f=a1++iaiXi1++dadXd1f^{\prime}=a_{1}+\cdots+ia_{i}X^{i-1}+\cdots+da_{d}X^{d-1}, which is the zero polynomial if only if FF has characteristic p0p\neq 0 and ai=0a_{i}=0 for all ii not divisible by pp.

(c) \Rightarrow (d). By hypothesis, f(X)=g(Xp)f(X)=g(X^{p}) with g(X)F[X]g(X)\in F[X]. Let g(X)=i(Xai)mig(X)=\prod_{i}(X-a_{i})^{m_{i}} in some extension field. Then each aia_{i} becomes a ppth power, say, ai=αipa_{i}=\alpha_{i}^{p}, in some possibly larger extension field. Now

f(X)=g(Xp)=i(Xpai)mi=i(Xαi)pmif(X)=g(X^{p})=\prod\nolimits_{i}(X^{p}-a_{i})^{m_{i}}=\prod\nolimits_{i}(X-% \alpha_{i})^{pm_{i}}

which shows that every root of f(X)f(X) has multiplicity at least pp.

(d) \Rightarrow (a). Obvious.

Proposition 2.13

The following conditions on a nonzero polynomial fF[X]f\in F[X] are equivalent:

  1. (a)

    gcd(f,f)=1\gcd(f,f^{\prime})=1 in F[X]F[X];

  2. (b)

    ff has only simple roots.

Proof.

Let Ω\Omega be an extension of FF splitting ff. From (5), p. 5, we see that a root α\alpha of ff in Ω\Omega is multiple if and only if it is also a root of ff^{\prime}.

If gcd(f,f)=1\gcd(f,f^{\prime})=1, then ff and ff^{\prime} have no common factor in Ω[X]\Omega[X] (see 2.10). In particular, they have no common root, and so ff has only simple roots.

If ff has only simple roots, then gcd(f,f)\gcd(f,f^{\prime}) must be the constant polynomial, because otherwise it would have a root in Ω\Omega which would then be a common root of ff and ff^{\prime}.

Definition 2.14

A polynomial is separable if it is nonzero and satisfies the equivalent conditions on (2.13).11 1 This is Bourbaki’s definition. Often (e.g., in the books of Jacobson and in earlier versions of these notes) a polynomial ff is said to be separable if each of its irreducible factors has only simple roots.

Thus a nonconstant irreducible polynomial ff is not separable if and only if FF has characteristic p0p\neq 0 and ff is a polynomial in XpX^{p} (see 2.12). Let f=fif=\prod f_{i} with ff and the fif_{i} monic and the fif_{i} irreducible; then ff is separable if and only if the fif_{i} are distinct and separable. If ff is separable as a polynomial in F[X]F[X], then it is separable as a polynomial in E[X]E[X] for every extension EE of FF.

Definition 2.15

A field FF is perfect if it has characteristic zero or it has characteristic pp and every every element of FF is a ppth power.

Thus, FF is perfect if and only if F=FqF=F^{q}, where qq is the characteristic exponent of FF.

Proposition 2.16

A field FF is perfect if and only if every irreducible polynomial in F[X]F[X] is separable.

Proof.

If FF has characteristic zero, the statement is obvious, and so we may suppose FF has characteristic p0p\neq 0. If FF contains an element aa that is not a ppth power, then XpaX^{p}-a is irreducible in F[X]F[X] but not separable (see 2.11). Conversely, if every element of FF is a ppth power, then every polynomial in XpX^{p} with coefficients in FF is a ppth power in F[X]F[X],

aiXip=(biXi)pif ai=bip,\textstyle\sum a_{i}X^{ip}=\left(\sum b_{i}X^{i}\right)^{p}\quad\text{if }\quad a_{i}=b_{i}^{p}\text{,}

and so it is not irreducible.

Example 2.17
  1. (a)

    A finite field FF is perfect, because the Frobenius endomorphism aap:FFa\mapsto a^{p}\colon F\rightarrow F is injective and therefore surjective (by counting).

  2. (b)

    A field that can be written as a union of perfect fields is perfect. Therefore, every field algebraic over 𝔽p\mathbb{F}{}_{p} is perfect.

  3. (c)

    Every algebraically closed field is perfect.

  4. (d)

    If F0F_{0} has characteristic p0p\neq 0, then F=F0(X)F=F_{0}(X) is not perfect, because XX is not a ppth power.

Aside 2.18

Let FF be a perfect field. We’ll see later (LABEL:ag1) that every finite extension E/FE/F is simple, i.e., E=F[α]E=F[\alpha] with α\alpha a root of a (separable) polynomial fF[X]f\in F[X] of degree [E:F][E\colon F]. Thus it follows directly from (2.2b) that, for any extension Ω\Omega of FF, the number of FF-homomorphisms EΩE\rightarrow\Omega is [E:F]\leq[E\colon F], with equality if and only if ff splits in Ω\Omega. We can’t use this argument here because it would make the exposition circular.