math.wiki
Abstract algebra › Fields › Splitting fields
Sign in

Splitting fields

The polynomial x^3 - 6 has one real root and two that are not real. What is the smallest field holding all three, how large is it, and does the answer depend on how we build it?

Where you are 5 in Fields

Why this exists

Kronecker’s theorem (Field extensions and degree, Corollary 4) gives a nonconstant polynomial one root in a larger field. Questions about all the roots at once, such as which field maps exchange them, need a field holding every root and nothing more. Is there a smallest such field, and how large is it? And since building it involves choices (which factor, which root, which model), do different choices give different fields? The answers (yes, of degree at most n!; and no, up to an isomorphism fixing the base field) are where Galois theory starts.

Definition

In plain words: a polynomial splits over a field when it is a product of linear factors there, and a splitting field is a field where it splits that its roots generate.

Let F be a field and f \in F[x] of degree n \ge 1 with leading coefficient c.

  • For an extension E/F, f splits over E if f = c(x - \alpha_1)(x - \alpha_2)\cdots(x - \alpha_n) in E[x] for some \alpha_1, \dots, \alpha_n \in E, not necessarily distinct.
  • An extension E/F is a splitting field of f over F if f splits over E as above and E = F(\alpha_1, \dots, \alpha_n).
  • A finite extension E/F is normal if every irreducible polynomial in F[x] with a root in E splits over E.
  • For a field isomorphism \theta : F \to F' and g = a_mx^m + \dots + a_0 \in F[x], write g^\theta = \theta(a_m)x^m + \dots + \theta(a_0) \in F'[x]. Polynomial sums and products are computed from those of coefficients, so g \mapsto g^\theta is a ring isomorphism F[x] \to F'[x] (Polynomial rings and the division algorithm); it keeps degrees and carries irreducible polynomials to irreducible ones.

For example, x^2 + 1 = (x - i)(x + i) over \mathbb{C}; \mathbb{Q}(i) is a splitting field of it over \mathbb{Q}, and \mathbb{C} is one over \mathbb{R}. A polynomial that already splits over F has F as its splitting field.

Intuition

The splitting field is everything computable from the roots and F with the four operations. Built one root at a time, each new root has degree at most that of what is left of the polynomial, which gives the bound n!. Two builds match because an isomorphism may send a root to any root of the matching polynomial, and Lemma 3 shows every such choice extends.

Re Im
\tfrac{2\pi}{3}
\displaystyle \sqrt[3]{6}
\displaystyle \sqrt[3]{6}\,\omega
\displaystyle \sqrt[3]{6}\,\omega^2
0
Figure 1The roots of x^3 - 6, with \displaystyle \omega = \operatorname{cis}\frac{2\pi}{3}. Only one is real, so \displaystyle \mathbb{Q}(\sqrt[3]{6}) \subseteq \mathbb{R} misses the other two.

Results

Throughout, F and F' are fields and f \in F[x] has degree n \ge 1 and leading coefficient c.

Theorem 1 (existence and size). (a) f has a splitting field over F. (b) Every splitting field E of f over F is finite over F, with [E : F] \le n!.

Lemma 2 (inside one field). Let L/F be an extension over which f = c(x - \alpha_1)\cdots(x - \alpha_n). Then (a) the roots of f in L are exactly \alpha_1, \dots, \alpha_n; (b) F(\alpha_1, \dots, \alpha_n) is a splitting field of f over F and lies inside every field K with F \subseteq K \subseteq L over which f splits, so it is the only splitting field of f over F inside L; (c) if E is a splitting field of f over F and F \subseteq K \subseteq E, then E is a splitting field of f over K.

Lemma 3 (extending by one root). Let \theta : F \to F' be an isomorphism, p \in F[x] irreducible, \alpha a root of p in an extension of F, and \beta a root of p^\theta in an extension of F'. Then there is exactly one isomorphism \psi : F(\alpha) \to F'(\beta) with \psi(a) = \theta(a) for every a \in F and \psi(\alpha) = \beta.

Theorem 4 (the isomorphism extension theorem). Let \theta : F \to F' be an isomorphism, E a splitting field of f over F, and E' a splitting field of f^\theta over F'. Then (a) \theta extends to an isomorphism \psi : E \to E'; (b) there are at most [E : F] such extensions, and exactly [E : F] if the n roots of f in E are distinct.

Corollary 5 (uniqueness). Any two splitting fields of f over F are isomorphic by an isomorphism fixing every element of F. If E is one, there are at most [E : F] isomorphisms E \to E fixing F elementwise, and exactly [E : F] if the n roots of f in E are distinct.

Theorem 6 (splitting fields are normal). A finite extension E/F is a splitting field over F of some nonconstant polynomial in F[x] if and only if it is normal.

Why it is true

Below, “FE” refers to Field extensions and degree and “AE” to Algebraic elements and minimal polynomials.

Proof of Lemma 2. (a) If f(\beta) = 0 with \beta \in L, then c(\beta - \alpha_1)\cdots(\beta - \alpha_n) = 0 in the field L with c \ne 0, so \beta = \alpha_i for some i. (b) The factorization has its coefficients in E_0 = F(\alpha_1, \dots, \alpha_n), so f splits over E_0, which its roots generate. If f = c(x - \beta_1)\cdots(x - \beta_n) with every \beta_j \in K, then (a), applied to both factorizations, says the \alpha_i and the \beta_j are the same elements, so E_0 \subseteq K. (c) f splits over E, and E = F(\alpha_1, \dots, \alpha_n) \subseteq K(\alpha_1, \dots, \alpha_n) \subseteq E. \square

Proof of Theorem 1. By induction on n, for all fields at once (Proof by induction). For n = 1, f = c(x - \alpha_1) with \alpha_1 = -a_0/c \in F, so F is a splitting field, and any splitting field is F(\alpha_1) = F. Let n \ge 2.

(a) By Kronecker’s theorem f has a root \alpha_1 in some extension of F; let K = F(\alpha_1). By the The remainder and factor theorems, f = (x - \alpha_1)g with g \in K[x] of degree n - 1 and leading coefficient c. By induction g has a splitting field E = K(\alpha_2, \dots, \alpha_n) over K, with g = c(x - \alpha_2)\cdots(x - \alpha_n). Then f splits over E = F(\alpha_1, \dots, \alpha_n) (FE Lemma 2).

(b) Let E = F(\alpha_1, \dots, \alpha_n) be a splitting field and K = F(\alpha_1). The minimal polynomial of \alpha_1 divides f (AE Theorem 1), so [K : F] \le n. Dividing f by x - \alpha_1 in K[x] (Polynomial long division) gives a quotient g \in K[x], which by uniqueness of the quotient in E[x] is c(x - \alpha_2)\cdots(x - \alpha_n). So E = K(\alpha_2, \dots, \alpha_n) is a splitting field of g over K, [E : K] \le (n - 1)! by induction, and the tower law (FE Theorem 6) gives [E : F] \le n \cdot (n - 1)! = n!. \square

Proof of Lemma 3. As g \mapsto g^\theta is a ring isomorphism, p \mid g - h exactly when p^\theta \mid g^\theta - h^\theta; so g + \langle p \rangle \mapsto g^\theta + \langle p^\theta \rangle is a well-defined, injective, onto ring homomorphism F[x]/\langle p \rangle \to F'[x]/\langle p^\theta \rangle. As p and p^\theta are irreducible, FE Theorem 5 gives isomorphisms g + \langle p \rangle \mapsto g(\alpha) onto F(\alpha) and h + \langle p^\theta \rangle \mapsto h(\beta) onto F'(\beta). Composing, \psi(g(\alpha)) = g^\theta(\beta) is an isomorphism sending each constant a to \theta(a) and \alpha to \beta. It is the only one: every element of F(\alpha) = F[\alpha] is g(\alpha) with g = \sum a_kx^k, and a ring homomorphism with the required values sends it to \sum \theta(a_k)\beta^k = g^\theta(\beta). \square

Proof of Theorem 4. By induction on [E : F], finite by Theorem 1, for all \theta at once. Write f = c(x - \alpha_1)\cdots(x - \alpha_n) in E[x] and f^\theta = \theta(c)(x - \beta_1)\cdots(x - \beta_n) in E'[x]. Any \psi extending \theta preserves sums and products, so \psi(g(\gamma)) = g^\theta(\psi(\gamma)) for g \in F[x] and \gamma \in E.

If [E : F] = 1, every \alpha_i \in F, so f^\theta = \theta(c)\prod_i (x - \theta(\alpha_i)) splits over F', and Lemma 2(b) in L = E' gives E' = F'. Then \psi = \theta is the only extension.

If [E : F] > 1, some \alpha = \alpha_i lies outside F. Let p = m_{\alpha, F}, of degree d = [F(\alpha) : F] \ge 2 (AE Theorem 1). As p \mid f, p^\theta \mid f^\theta, so by unique factorization in E'[x] (Unique factorization domains) p^\theta is a product of d of the factors x - \beta_j, with r distinct roots in E', 1 \le r \le d. For each such root \beta, Lemma 3 gives one isomorphism \theta_\beta : F(\alpha) \to F'(\beta) extending \theta with \alpha \mapsto \beta. By Lemma 2(c), E is a splitting field of f over F(\alpha) and E' one of f^{\theta_\beta} = f^\theta over F'(\beta), and [E : F(\alpha)] = [E : F]/d < [E : F]. By induction each \theta_\beta extends to E, in at most [E : F]/d ways. Every extension \psi of \theta sends \alpha to a root of p^\theta, as p^\theta(\psi(\alpha)) = \psi(p(\alpha)) = 0, and agrees on F(\alpha) with \theta_{\psi(\alpha)} by the uniqueness in Lemma 3. So the extensions of \theta are exactly those of the \theta_\beta: at least one, and at most r \cdot [E : F]/d \le [E : F].

Now let the \alpha_i be distinct. By (a) there is an isomorphism \psi_0 : E \to E' extending \theta, and f^\theta = \theta(c)\prod_i (x - \psi_0(\alpha_i)), so by Lemma 2(a) the \beta_j are the distinct \psi_0(\alpha_i). Then r = d, the roots stay distinct over F(\alpha), and by induction (base case 1 = [E : F]) each \theta_\beta has exactly [E : F]/d extensions: [E : F] in all. \square

Proof of Corollary 5. Take F' = F and \theta = \mathrm{id}_F, so that f^\theta = f. \square

Proof of Theorem 6. Let E = F(\alpha_1, \dots, \alpha_n) be a splitting field of f and p \in F[x] irreducible with a root \alpha \in E. Let L be a splitting field of p over E (Theorem 1) and \beta \in L any root of p. Lemma 3 gives an isomorphism F(\alpha) \to F(\beta) fixing F. By Lemma 2(c), E is a splitting field of f over F(\alpha), and E(\beta) = F(\beta)(\alpha_1, \dots, \alpha_n) is one over F(\beta), so Theorem 4(a) gives an isomorphism E \to E(\beta) fixing F. It is F-linear, carrying a basis to a basis, so [E : F] = [E(\beta) : F], and the tower law gives [E(\beta) : E] = 1: \beta \in E. So p, which splits over L, has all its roots in E and splits over E.

Conversely, let E/F be finite and normal. Then E = F(\gamma_1, \dots, \gamma_k), k \ge 1, with each \gamma_j algebraic over F (AE Corollary 6). Each m_{\gamma_j} is irreducible with a root in E, so it splits over E, and so does f = m_{\gamma_1}\cdots m_{\gamma_k}, whose roots include the \gamma_j and so generate E. \square

Worked examples

Example 1

Two quadratic factors

Find a splitting field E of f = x^4 - x^2 - 6 over \mathbb{Q}, its degree, and the isomorphisms E \to E fixing \mathbb{Q}.

Solution

As a quadratic in x^2, f = (x^2 - 3)(x^2 + 2), with roots \displaystyle \pm\sqrt{3} and \displaystyle \pm\sqrt{2}\,i. By Lemma 2(b), \displaystyle E = \mathbb{Q}(\sqrt{3}, \sqrt{2}\,i). Now \displaystyle [\mathbb{Q}(\sqrt{3}) : \mathbb{Q}] = 2 (Irrational numbers), and \displaystyle \sqrt{2}\,i \notin \mathbb{Q}(\sqrt{3}) \subseteq \mathbb{R}, so x^2 + 2 is irreducible over \displaystyle \mathbb{Q}(\sqrt{3}). The tower law gives [E : \mathbb{Q}] = 2 \cdot 2 = 4, well below 4! = 24.

An isomorphism \sigma : E \to E fixing \mathbb{Q} sends a root of x^2 - 3 to a root of x^2 - 3, so \displaystyle \sigma(\sqrt{3}) = \pm\sqrt{3}; likewise \displaystyle \sigma(\sqrt{2}\,i) = \pm\sqrt{2}\,i, and these values determine \sigma. That allows at most 4 maps, and as the four roots of f are distinct, Corollary 5 gives exactly 4: every choice of signs occurs. One is complex conjugation.

Example 2

A real root is not enough

Let f = x^3 - 6. Find a splitting field of f over \mathbb{Q} and its degree. Show that the three fields generated by one root each are isomorphic but distinct, and find the degree of a splitting field of f over \displaystyle \mathbb{Q}(\sqrt{-3}) and over \mathbb{R}.

Solution

Let \displaystyle a = \sqrt[3]{6} and \displaystyle \omega = \operatorname{cis}\frac{2\pi}{3} = \frac{-1 + \sqrt{3}\,i}{2}, a root of x^2 + x + 1 (Roots of a complex number and roots of unity). The roots of f are a, a\omega, a\omega^2, and E = \mathbb{Q}(a, a\omega, a\omega^2) contains \omega = a\omega/a, so E = \mathbb{Q}(a, \omega). Eisenstein’s criterion at 2 (Irreducible polynomials and Eisenstein's criterion) makes f irreducible, so [\mathbb{Q}(a) : \mathbb{Q}] = 3. As \mathbb{Q}(a) \subseteq \mathbb{R} and \omega \notin \mathbb{R}, x^2 + x + 1 is irreducible over \mathbb{Q}(a), and [E : \mathbb{Q}] = 3 \cdot 2 = 6 = 3!.

Lemma 3, with \theta = \mathrm{id}_{\mathbb{Q}} and p = f, gives an isomorphism \mathbb{Q}(a) \to \mathbb{Q}(a\omega^k) fixing \mathbb{Q} with a \mapsto a\omega^k. Yet if \mathbb{Q}(a\omega^j) = \mathbb{Q}(a\omega^k) with j \ne k, this field of degree 3 would contain the quotient \omega^{k - j}, which is \omega or \omega^{-1}, hence \omega, of degree 2; and 2 \nmid 3 (FE Corollary 7(b)).

Since \displaystyle \sqrt{-3} = 2\omega + 1, \displaystyle \mathbb{Q}(\omega) = \mathbb{Q}(\sqrt{-3}), and by Lemma 2(c) E is a splitting field of f over \displaystyle \mathbb{Q}(\sqrt{-3}), of degree 6/2 = 3. Over \mathbb{R}, f = (x - a)(x^2 + ax + a^2), and the splitting field \mathbb{R}(\omega) = \mathbb{C} has degree 2.

2
2
2
3
3
3
3
2
\displaystyle E = \mathbb{Q}(\sqrt[3]{6}, \omega)
\displaystyle \mathbb{Q}(\sqrt[3]{6})
\displaystyle \mathbb{Q}(\sqrt[3]{6}\,\omega)
\displaystyle \mathbb{Q}(\sqrt[3]{6}\,\omega^2)
\displaystyle \mathbb{Q}(\sqrt{-3})
\mathbb{Q}
Figure 2Example 2: the splitting field E of x^3 - 6 over \mathbb{Q} and four fields between. The three cubic fields are isomorphic (Lemma 3) but distinct, and every path from bottom to top multiplies to [E : \mathbb{Q}] = 6.
Example 3

One root is enough over \mathbb{Z}_3

Writing k for [k], let f = x^3 - x + 1 \in \mathbb{Z}_3[x]. Show that K = \mathbb{Z}_3[x]/\langle f \rangle is a splitting field of f over \mathbb{Z}_3, and count the isomorphisms K \to K.

Solution

At 0, 1, 2, f takes 1, 1, 7 = 1; a cubic with no root is irreducible, so K is a field of degree 3 containing the root \alpha = x + \langle f \rangle (FE Theorem 3). Expanding,

\begin{aligned} f(x + 1) &= (x^3 + 3x^2 + 3x + 1) - (x + 1) + 1 \\ &= f(x) + 3x^2 + 3x, \end{aligned}

and 3 = 0 in \mathbb{Z}_3, so f(x + 1) = f(x) in \mathbb{Z}_3[x]. Hence f(\alpha + 1) = f(\alpha) = 0 and f(\alpha + 2) = f(\alpha + 1) = 0. The distinct roots split off one at a time by the factor theorem, so f = (x - \alpha)(x - \alpha - 1)(x - \alpha - 2) in K[x], and they generate K = \mathbb{Z}_3(\alpha): K is a splitting field, of degree 3 rather than 3!. Every isomorphism K \to K sends 1 to 1, so it fixes \mathbb{Z}_3, and Corollary 5 gives exactly 3: \alpha \mapsto \alpha + k for k = 0, 1, 2.

Example 4

One root generates them all

Find a splitting field of f = x^4 - 4x^2 + 2 over \mathbb{Q} and its degree.

Solution

Eisenstein at 2 makes f irreducible. Solving for x^2 gives \displaystyle x^2 = 2 \pm \sqrt{2}, both positive, so the roots are \pm\alpha and \pm\beta with \displaystyle \alpha = \sqrt{2 + \sqrt{2}} and \displaystyle \beta = \sqrt{2 - \sqrt{2}}. Now \displaystyle \sqrt{2} = \alpha^2 - 2 and \displaystyle \alpha\beta = \sqrt{(2 + \sqrt{2})(2 - \sqrt{2})} = \sqrt{2}. From f(\alpha) = 0, \alpha(\alpha^3 - 4\alpha) = -2, so \displaystyle \alpha^{-1} = \frac{1}{2}(4\alpha - \alpha^3) and

\beta = \frac{\alpha^2 - 2}{\alpha} = \alpha - 2\alpha^{-1} = \alpha^3 - 3\alpha.

All four roots lie in \mathbb{Q}(\alpha), so E = \mathbb{Q}(\alpha) and [E : \mathbb{Q}] = 4, the least possible for an irreducible quartic.

Example 5

Which extensions are splitting fields

Prove that every extension of degree 2 is a splitting field, and that \displaystyle \mathbb{Q}(\sqrt[3]{6}) is not a splitting field over \mathbb{Q} of any polynomial.

Solution

Proof. Let [E : F] = 2 and \alpha \in E \setminus F. Then E = F(\alpha) (FE Corollary 7(c)), and m_\alpha = x^2 + bx + c with b, c \in F. Put \alpha' = -b - \alpha \in E. Since \alpha^2 + b\alpha = -c,

\begin{aligned} (x - \alpha)(x - \alpha') &= x^2 + bx - (\alpha^2 + b\alpha) \\ &= x^2 + bx + c, \end{aligned}

so m_\alpha splits over E = F(\alpha, \alpha'), in any characteristic. If \displaystyle \mathbb{Q}(\sqrt[3]{6}) were a splitting field over \mathbb{Q}, it would be normal (Theorem 6), so the irreducible x^3 - 6, with the root \displaystyle \sqrt[3]{6} in it, would split over it. But its roots \displaystyle \sqrt[3]{6}\,\omega and \displaystyle \sqrt[3]{6}\,\omega^2 are not real. \square

Boundaries

  • Unique up to isomorphism, not more. Inside one field L there is only one splitting field (Lemma 2(b)), but the isomorphism of Corollary 5 is not unique: Example 1 has four. Studying all of them is the work of Field automorphisms and the Galois group.
  • Repeated roots lower the count. Over F = \mathbb{Z}_2(t), the rational functions over \mathbb{Z}_2, f = x^2 - t is irreducible (g^2 = th^2 is impossible by parity of degrees). In a splitting field F(s) with s^2 = t, f = (x - s)^2 since 2 = 0, so the degree is 2 but only one isomorphism fixes F, as s must go to a root of f. Over \mathbb{Q} and finite fields an irreducible polynomial has distinct roots, as Field automorphisms and the Galois group shows.
  • Infinitely many polynomials. A field over which every polynomial in F[x] splits, generated by their roots, is an algebraic closure of F; its existence uses Zorn’s lemma and is beyond this course (Algebraic elements and minimal polynomials).

Where this goes next

Finite fields builds the field with p^k elements as a splitting field of x^{p^k} - x over \mathbb{Z}_p, and Corollary 5 is why there is only one. Field automorphisms and the Galois group makes the isomorphisms E \to E fixing F into a group, the Galois group, of order [E : F] when the roots are distinct; by Exercise 13 it permutes the roots, so then [E : F] divides n! (Lagrange's theorem). Normality (Theorem 6) is the hypothesis the The fundamental theorem of Galois theory runs on.

Exercises

Splitting fields over \mathbb{Q}

  1. Find a splitting field of x^4 - 7x^2 + 10 over \mathbb{Q} and its degree.
  2. Find a splitting field of x^6 - 1 over \mathbb{Q} and its degree.
  3. Find a splitting field of x^4 + 4 over \mathbb{Q} and its degree.
  4. Find a splitting field of x^4 - 6x^2 + 4 over \mathbb{Q} and its degree, and compare it with Exercise 1.
  5. Show that x^4 - 2x^2 - 1 is irreducible over \mathbb{Q} and that its splitting field over \mathbb{Q} has degree 8.

Splitting fields over \mathbb{Z}_3

Writing k for [k]:

  1. Find a splitting field over \mathbb{Z}_3, and its degree, of x^2 + 2 and of x^2 + 2x + 2.
  2. Factor x^4 + 1 into irreducibles in \mathbb{Z}_3[x] and show that its splitting field over \mathbb{Z}_3 has degree 2.

Isomorphisms

  1. Let \displaystyle E = \mathbb{Q}(\sqrt{2}, \sqrt{5}). How many isomorphisms E \to E fix \mathbb{Q}? Describe each by its values at \displaystyle \sqrt{2} and \displaystyle \sqrt{5}.
  2. After a colloquium, Isaac builds the field \mathbb{Q}[x]/\langle x^2 + 3 \rangle and Leah builds \mathbb{Q}[x]/\langle x^2 + x + 1 \rangle. Show that their fields are isomorphic by an isomorphism fixing \mathbb{Q}, and write one down.

Proofs

  1. Let f \in F[x] be an irreducible cubic with splitting field E. Prove that [E : F] is 3 or 6, and give an example of each.
  2. Let f \in F[x] have degree n \ge 2 and a splitting field E with [E : F] = n!. Prove that f is irreducible over F.
  3. Writing k for [k], prove that f = x^5 - x - 1 is irreducible over \mathbb{Z}_5 and that its splitting field has degree 5. (Compare f(x + 1) with f(x).)
  4. Let E = F(\alpha_1, \dots, \alpha_n) be a splitting field of f over F and \sigma : E \to E an isomorphism fixing F elementwise. Prove that \sigma permutes the distinct roots of f and is determined by that permutation.
  5. Refute with a counterexample: if F \subseteq K \subseteq E, K is a splitting field of some polynomial over F, and E is a splitting field of some polynomial over K, then E is a splitting field of some polynomial over F.
Answers
  1. x^4 - 7x^2 + 10 = (x^2 - 2)(x^2 - 5), so \displaystyle E = \mathbb{Q}(\sqrt{2}, \sqrt{5}). If \displaystyle \sqrt{5} = a + b\sqrt{2} with a, b \in \mathbb{Q}, then \displaystyle 5 = a^2 + 2b^2 + 2ab\sqrt{2} forces ab = 0; b = 0 makes \displaystyle \sqrt{5} rational, and a = 0 gives (2b)^2 = 10, both false. So [E : \mathbb{Q}] = 2 \cdot 2 = 4.
  2. x^6 - 1 is (x - 1)(x + 1) times (x^2 + x + 1)(x^2 - x + 1), with roots \pm 1 and \displaystyle \frac{\pm 1 \pm \sqrt{3}\,i}{2}. They generate \displaystyle E = \mathbb{Q}(\sqrt{3}\,i), of degree 2.
  3. x^4 + 4 = (x^2 + 2)^2 - 4x^2, which is (x^2 + 2x + 2)(x^2 - 2x + 2), with roots \pm 1 \pm i. So E = \mathbb{Q}(i), of degree 2, not 4.
  4. \displaystyle x^2 = 3 \pm \sqrt{5}, and \displaystyle \big(\frac{\sqrt{10} \pm \sqrt{2}}{2}\big)^2 = \frac{12 \pm 4\sqrt{5}}{4} = 3 \pm \sqrt{5}, so the roots are \displaystyle \pm\frac{\sqrt{10} \pm \sqrt{2}}{2}. Their sums and differences give \displaystyle \sqrt{10} and \displaystyle \sqrt{2}, so \displaystyle E = \mathbb{Q}(\sqrt{2}, \sqrt{10}) = \mathbb{Q}(\sqrt{2}, \sqrt{5}), as \displaystyle \sqrt{10} = \sqrt{2}\sqrt{5}: the field of Exercise 1, of degree 4.
  5. Let \displaystyle \alpha = \sqrt{1 + \sqrt{2}}, a root since (\alpha^2 - 1)^2 = 2. If \displaystyle \alpha = a + b\sqrt{2} with a, b \in \mathbb{Q}, then 2ab = 1 and a^2 + 2b^2 = 1, so 2a^4 - 2a^2 + 1 = 0, impossible for real a (4 - 8 < 0). So \displaystyle \alpha \notin \mathbb{Q}(\sqrt{2}) = \mathbb{Q}(\alpha^2), [\mathbb{Q}(\alpha) : \mathbb{Q}] = 2 \cdot 2 = 4, and the quartic is m_\alpha, irreducible. The other roots are -\alpha and \pm\beta with \displaystyle (\alpha\beta)^2 = (1 + \sqrt{2})(1 - \sqrt{2}) = -1, so \beta = \pm i/\alpha and E = \mathbb{Q}(\alpha, i). As \mathbb{Q}(\alpha) \subseteq \mathbb{R}, [E : \mathbb{Q}] = 4 \cdot 2 = 8.
  6. x^2 + 2 = x^2 - 1 = (x - 1)(x + 1) splits over \mathbb{Z}_3, its own splitting field, of degree 1. The quadratic x^2 + 2x + 2 takes 2, 2, 1 at 0, 1, 2, so it is irreducible; in K = \mathbb{Z}_3[x]/\langle x^2 + 2x + 2 \rangle with root \gamma, the other root is -2 - \gamma = 1 - \gamma (Example 5), so K is a splitting field, of degree 2.
  7. (x^2 + x + 2)(x^2 + 2x + 2) = x^4 + 3x^3 + 6x^2 + 6x + 4 = x^4 + 1 in \mathbb{Z}_3[x]. Neither factor has a root (x^2 + x + 2 takes 2, 1, 2; the other, Exercise 6), so both are irreducible. In K = \mathbb{Z}_3[x]/\langle x^2 + x + 2 \rangle with root \alpha, the first factor has roots \alpha, 2 + 2\alpha and the second 2\alpha, 1 + \alpha, since (-\alpha)^2 + 2(-\alpha) + 2 = \alpha^2 + \alpha + 2 = 0. So x^4 + 1 splits over K = \mathbb{Z}_3(\alpha), of degree 2.
  8. Four. Such a map sends \displaystyle \sqrt{2} to a root of x^2 - 2 and \displaystyle \sqrt{5} to a root of x^2 - 5, and these values determine it, so there are at most 4. E is the splitting field of Exercise 1, with four distinct roots, so Corollary 5 gives exactly 4: \displaystyle \sqrt{2} \mapsto \pm\sqrt{2}, \displaystyle \sqrt{5} \mapsto \pm\sqrt{5}, all four sign choices.
  9. Neither polynomial has a rational root, so both quotients are fields of degree 2. In Isaac’s, \gamma = x + \langle x^2 + 3 \rangle and -\gamma are the roots of x^2 + 3, so it is a splitting field of x^2 + 3. In Leah’s, with \beta = x + \langle x^2 + x + 1 \rangle, (2\beta + 1)^2 = 4(\beta^2 + \beta + 1) - 3 = -3, so \pm(2\beta + 1) are the roots of x^2 + 3, and they generate, as \displaystyle \beta = \frac{(2\beta + 1) - 1}{2}. Lemma 3, with \theta = \mathrm{id}_{\mathbb{Q}} and \gamma \mapsto 2\beta + 1, gives one fixing \mathbb{Q}: a + b\gamma \mapsto a + b(2\beta + 1).
  10. Proof. Let \alpha \in E be a root; f is irreducible, so [F(\alpha) : F] = 3 (AE Theorem 1). In F(\alpha)[x], f = (x - \alpha)g with g quadratic, and as in the proof of Theorem 1(b), E is a splitting field of g over F(\alpha), so [E : F(\alpha)] \le 2! = 2. By the tower law, [E : F] is 3 or 6. \square Example 3 has degree 3, Example 2 degree 6.
  11. Proof. Suppose f = gh with 1 \le k = \deg g \le n - 1. In E[x], f = c(x - \alpha_1)\cdots(x - \alpha_n), so by unique factorization in E[x], after renumbering, g = c_1(x - \alpha_1)\cdots(x - \alpha_k) and h = c_2(x - \alpha_{k+1})\cdots(x - \alpha_n). Then K = F(\alpha_1, \dots, \alpha_k) is a splitting field of g over F, and E = K(\alpha_{k+1}, \dots, \alpha_n) one of h over K. By Theorem 1(b) and the tower law, [E : F] \le k!\,(n - k)! = n!/\binom{n}{k}, and \binom{n}{k} \ge n \ge 2, so [E : F] < n!. \square
  12. Proof. For 1 \le j \le 4, 5 \mid \binom{5}{j}, so (x + 1)^5 = x^5 + 1 in \mathbb{Z}_5[x] and f(x + 1) = x^5 + 1 - x - 1 - 1 = f(x). Let E be a splitting field and \alpha \in E a root. Then \alpha, \alpha + 1, \dots, \alpha + 4 are five distinct roots, hence all of them (Lemma 2(a)), and E = \mathbb{Z}_5(\alpha). By unique factorization in E[x], each monic irreducible factor q of f in \mathbb{Z}_5[x] has some \alpha + k as a root, so q = m_{\alpha + k} (AE Theorem 1(d)) and \deg q = [\mathbb{Z}_5(\alpha + k) : \mathbb{Z}_5] = [E : \mathbb{Z}_5] = d. So d divides 5. By Fermat's little theorem, c^5 = c for c \in \mathbb{Z}_5, so f(c) = -1 \ne 0 and d \ne 1. Hence d = 5: f is irreducible and [E : \mathbb{Z}_5] = 5. \square
  13. Proof. For a root \alpha_i, f(\sigma(\alpha_i)) = \sigma(f(\alpha_i)) = 0, as \sigma preserves sums and products and fixes the coefficients; so \sigma(\alpha_i) is a root (Lemma 2(a)). \sigma is injective, so it maps the finite set of distinct roots onto itself. Adjoining one algebraic element at a time, E = F[\alpha_1, \dots, \alpha_n] (AE Theorem 1(e)), so every element is a polynomial expression in the \alpha_i with coefficients in F, and \sigma sends it to the same expression in the \sigma(\alpha_i). \square
  14. Take \displaystyle \mathbb{Q} \subseteq \mathbb{Q}(\sqrt{3}) \subseteq \mathbb{Q}(\sqrt[4]{3}). As x^4 - 3 is Eisenstein at 3 and \displaystyle (\sqrt[4]{3})^2 = \sqrt{3}, each step has degree 2, so each is a splitting field by Example 5. But x^4 - 3 is irreducible over \mathbb{Q} with the root \displaystyle \sqrt[4]{3} in \displaystyle \mathbb{Q}(\sqrt[4]{3}) \subseteq \mathbb{R}, while its root \displaystyle \sqrt[4]{3}\,i is not real; so \displaystyle \mathbb{Q}(\sqrt[4]{3}) is not normal over \mathbb{Q}, hence not a splitting field (Theorem 6).
math.wiki · updated 2026-11-18 · CC-BY-SA-4.0