math.wiki
Abstract algebra › Fields › Field extensions and degree
Sign in

Field extensions and degree

Adjoining \displaystyle \sqrt{3} to \mathbb{Q} gives the numbers \displaystyle a + b\sqrt{3}, each fixed by two rational coordinates. How do we measure how much larger one field is than another, and what happens to that measure when we adjoin one element after another?

Where you are 5 in Fields

Why this exists

The field \mathbb{Q} has no square root of 3, and x^3 - 2x - 2 has no rational root. Inside \mathbb{C} we can adjoin such a number and take the smallest field containing it; over \mathbb{Z}_p, with no \mathbb{C} around, we can still build a field holding a root, as Prime and maximal ideals began to do with F[x]/\langle f \rangle. Either way two fields F \subseteq E appear, and linear algebra measures the gap: E is a vector space over F, and its dimension, the degree, multiplies along a chain of fields. That one rule settles questions such as whether \displaystyle \sqrt{3} can be written with \displaystyle \sqrt[3]{7} and rational numbers (it cannot, Example 3), and it drives Compass-and-straightedge constructions, and why an angle cannot be trisected and Galois theory.

Definition

In plain words: an extension is a pair of fields, one inside the other; the larger is a vector space over the smaller, and the degree is its dimension.

  • A subfield of a field E is a subset containing 0 and 1 and closed under addition, subtraction, multiplication and inverses of nonzero elements. A field extension E/F, read “E over F”, is a field E with a subfield F. A field K with F \subseteq K \subseteq E is an intermediate field, and F \subseteq K \subseteq E is a tower.
  • A ring homomorphism \iota : F \to E out of a field is injective, as its kernel is an ideal of F without 1, hence \{0\} (Ideals and quotient rings). We then identify F with \iota(F) and call E an extension of F too.
  • Under its own addition and the products cv with c \in F, E is a vector space over F (Theorem 1). The degree [E : F] is its dimension \dim_F E; the extension is finite when the degree is finite and infinite otherwise. The vector-space lessons write \mathbb{F} for the scalars; every result of Bases and dimension, proved used here holds over any field.
  • For S \subseteq E, F(S) is the intersection of all subfields of E containing F \cup S. Each closure condition passes to intersections, so F(S) is the smallest subfield containing F \cup S: “F adjoin S”, written F(\alpha), F(\alpha_1, \dots, \alpha_n). An extension F(\alpha)/F is simple. The ring F[\alpha] = \{g(\alpha) : g \in F[x]\} lies inside F(\alpha).

For example [\mathbb{C} : \mathbb{R}] = 2 with basis (1, i). The set \displaystyle \{a + b\sqrt{3} : a, b \in \mathbb{Q}\} is closed under the four operations, since \displaystyle (a + b\sqrt{3})^{-1} = (a - b\sqrt{3})/(a^2 - 3b^2) and a^2 \ne 3b^2 unless a = b = 0 (Irrational numbers); so it is \displaystyle \mathbb{Q}(\sqrt{3}), of degree 2 with basis \displaystyle (1, \sqrt{3}).

Intuition

A degree trades a hard question (which numbers lie in E?) for a count (how many coordinates over F does an element need?). Adjoining a root \alpha of an irreducible polynomial of degree n adds exactly n coordinates, on 1, \alpha, \dots, \alpha^{n-1}, because the polynomial rewrites \alpha^n through lower powers. Adjoining in stages multiplies the counts: each coordinate over the middle field is itself a list of coordinates over the bottom one.

2
3
3
2
\displaystyle \mathbb{Q}(\sqrt{3}, \sqrt[3]{7})
\displaystyle \mathbb{Q}(\sqrt{3})
\displaystyle \mathbb{Q}(\sqrt[3]{7})
\mathbb{Q}
Figure 1The tower law along both sides of Example 3. Each path from bottom to top multiplies to \displaystyle [\mathbb{Q}(\sqrt{3}, \sqrt[3]{7}) : \mathbb{Q}] = 6.

Results

Throughout, F, K, E are fields.

Theorem 1 (the degree). Let E/F be an extension. (a) E is a vector space over F. (b) [E : F] = 1 if and only if E = F. (c) If F has q elements and [E : F] = n, then E has q^n elements.

Lemma 2 (adjoining in steps). If E/F is an extension and \alpha, \beta \in E, then F(\alpha, \beta) = F(\alpha)(\beta); by induction, F(\alpha_1, \dots, \alpha_n) = F(\alpha_1, \dots, \alpha_{n-1})(\alpha_n).

Theorem 3 (Kronecker’s construction). Let f \in F[x] be irreducible of degree n and K = F[x]/\langle f \rangle. Then K is a field, c \mapsto c + \langle f \rangle is an injective homomorphism F \to K (making K an extension of F), and \alpha = x + \langle f \rangle is a root of f in K. The list (1, \alpha, \dots, \alpha^{n-1}) is a basis of K over F, so [K : F] = n, and K = F(\alpha).

Corollary 4 (Kronecker’s theorem). Every nonconstant f \in F[x] has a root in some extension K of F with [K : F] \le \deg f.

Theorem 5 (simple extensions). Let E/F be an extension, \alpha \in E, and f \in F[x] irreducible of degree n with f(\alpha) = 0. Then: (a) for g \in F[x], g(\alpha) = 0 if and only if f \mid g; (b) g + \langle f \rangle \mapsto g(\alpha) is an injective ring homomorphism F[x]/\langle f \rangle \to E with image F[\alpha]; (c) F[\alpha] = F(\alpha), the list (1, \alpha, \dots, \alpha^{n-1}) is a basis of F(\alpha) over F, and [F(\alpha) : F] = n.

Theorem 6 (the tower law). Let F \subseteq K \subseteq E. Then E/F is finite if and only if E/K and K/F are, and then

[E : F] = [E : K]\,[K : F].

If (a_1, \dots, a_m) is a basis of K over F and (b_1, \dots, b_n) one of E over K, the mn products a_ib_j form a basis of E over F.

Corollary 7. Let E/F be finite. (a) For every intermediate field K, [K : F] and [E : K] divide [E : F]. (b) If \alpha \in E is a root of an irreducible f \in F[x], then \deg f divides [E : F]. (c) If [E : F] is prime, the only intermediate fields are F and E, and E = F(\alpha) for every \alpha \in E \setminus F.

Corollary 8 (two generators). Let E/F be an extension and \alpha, \beta \in E roots of irreducible f, g \in F[x] of degrees m and n. Then [F(\alpha, \beta) : F] \le mn, and both m and n divide it. If \gcd(m, n) = 1, it equals mn.

Why it is true

“Result k” is Result k of Bases and dimension, proved; “Theorem k there” is Theorem k of Prime and maximal ideals.

Proof of Theorem 1. (a) The axioms of Vector spaces from axioms with scalars in F are field axioms of E: addition is an abelian group, and c(v + w) = cv + cw, (c + d)v = cv + dv, (cd)v = c(dv), 1v = v hold in E. (b) If [E : F] = 1, the independent list (1) has length \dim_F E, so it is a basis (Result 5) and every v \in E is c \cdot 1 \in F. Conversely (1) is a basis of F over F. (c) For a basis (e_1, \dots, e_n), coordinates give a bijection E \to F^n (Result 1), and F^n has q^n elements. \square

Proof of Lemma 2. F(\alpha)(\beta) is a subfield containing F, \alpha, \beta, so it contains F(\alpha, \beta). And F(\alpha, \beta) is a subfield containing F and \alpha, hence F(\alpha), and \beta, so it contains F(\alpha)(\beta). \square

Proof of Theorem 3. As f is irreducible, \langle f \rangle is maximal (Theorem 7(b) there), so K is a field (Theorem 3 there). The map c \mapsto c + \langle f \rangle is the inclusion F \to F[x] followed by the natural map, a ring homomorphism out of a field, so injective. The natural map preserves sums and products, so f(x + \langle f \rangle) = f + \langle f \rangle = 0. By Theorem 7(c) there, each coset holds exactly one r = c_0 + c_1x + \dots + c_{n-1}x^{n-1}, and r + \langle f \rangle = c_0 + c_1\alpha + \dots + c_{n-1}\alpha^{n-1}. So every element of K is a combination of 1, \alpha, \dots, \alpha^{n-1} in exactly one way, a basis by Result 1. As F(\alpha) contains F and \alpha, it contains every such combination, so F(\alpha) = K. \square

Proof of Corollary 4. Every nonconstant polynomial over a field is a product of irreducible ones, by Strong induction and well-ordering on the degree: degree 1 is irreducible, and a reducible f is gh with g, h of smaller positive degree. So f = pq with p irreducible. Theorem 3 gives K = F[x]/\langle p \rangle and a root \alpha of p; then f(\alpha) = p(\alpha)q(\alpha) = 0 and [K : F] = \deg p \le \deg f. \square

Proof of Theorem 5. Evaluation \varepsilon(g) = g(\alpha) is a ring homomorphism F[x] \to E, as E is commutative. (a) Its kernel J is an ideal, so J = \langle h \rangle (Theorem 7(a) there). As f \in J, f = hu; as \varepsilon(1) = 1, h is not constant; irreducibility of f makes u a nonzero constant, so J = \langle f \rangle. (b) g(\alpha) = g'(\alpha) exactly when g - g' \in J = \langle f \rangle, so the map is well defined and injective, a homomorphism because \varepsilon is, with image F[\alpha]. (c) By (b), F[\alpha] is an isomorphic copy of the field of Theorem 3, so a subfield containing F and \alpha; hence F(\alpha) \subseteq F[\alpha] \subseteq F(\alpha). The isomorphism fixes F, so it is F-linear and carries the basis of Theorem 3 to (1, \alpha, \dots, \alpha^{n-1}). \square

Proof of Theorem 6. Let K/F and E/K be finite with the bases named. Each v \in E is \sum_j c_jb_j with c_j \in K, and c_j = \sum_i d_{ij}a_i with d_{ij} \in F, so v = \sum_{i,j} d_{ij}a_ib_j. If \sum_{i,j} d_{ij}a_ib_j = 0, then \sum_j \big(\sum_i d_{ij}a_i\big)b_j = 0 with coefficients in K, so each inner sum is 0 by independence of the b_j, and each d_{ij} = 0 by independence of the a_i.

Conversely let E/F be finite. K is an F-subspace of E, so finite-dimensional (Result 6). An F-basis of E spans E over K \supseteq F, so it shrinks to a K-basis (Result 2). \square

Proof of Corollary 7. (a) is Theorem 6. (b) F \subseteq F(\alpha) \subseteq E with [F(\alpha) : F] = \deg f (Theorem 5); apply (a). (c) [K : F] divides the prime [E : F]. If it is 1, K = F; otherwise [E : K] = 1 and K = E (Theorem 1(b)). For \alpha \notin F, F(\alpha) \ne F, so F(\alpha) = E. \square

Proof of Corollary 8. Let L = F(\alpha), of degree m. In L[x], g is a product of irreducibles (proof of Corollary 4), and g(\beta) = 0 in the field E, so \beta is a root of an irreducible factor h with \deg h \le n. By Theorem 5 over L and Lemma 2, [F(\alpha, \beta) : L] = \deg h, so [F(\alpha, \beta) : F] = m \deg h \le mn (Theorem 6). Corollary 7(b) gives m and n as divisors. If \gcd(m, n) = 1, then mn divides it, forcing equality. \square

Worked examples

Example 1

Computing in Kronecker’s field

Let f = x^3 - 2x - 2 \in \mathbb{Q}[x], K = \mathbb{Q}[x]/\langle f \rangle and \alpha = x + \langle f \rangle. Show that [K : \mathbb{Q}] = 3, write \alpha^4 and (\alpha + 1)^{-1} in the basis (1, \alpha, \alpha^2), and relate K to the roots of f in \mathbb{C}.

Solution

By the The rational root theorem a rational root would be \pm 1 or \pm 2, where f takes -3, -1, 2, -6. A cubic with no root is irreducible (Prime and maximal ideals), so Theorem 3 makes K a field with basis (1, \alpha, \alpha^2).

In K, f(\alpha) = 0 says \alpha^3 = 2\alpha + 2, so \alpha^4 = \alpha \cdot \alpha^3 = 2\alpha^2 + 2\alpha. Dividing (Polynomial long division), f = (x + 1)(x^2 - x - 1) - 1; at \alpha this reads (\alpha + 1)(\alpha^2 - \alpha - 1) = 1, so (\alpha + 1)^{-1} = \alpha^2 - \alpha - 1.

In \mathbb{C}: f' = 3x^2 - 2, and the local maximum \displaystyle f(-\sqrt{2/3}) = \frac{4}{3}\sqrt{2/3} - 2 is negative, so f has one real root r \approx 1.769 and two non-real roots z, \bar z (Complex roots come in pairs). By Theorem 5, \mathbb{Q}(r), \mathbb{Q}(z), \mathbb{Q}(\bar z) are each isomorphic to K, yet \mathbb{Q}(r) \subseteq \mathbb{R} while z \notin \mathbb{R}. One abstract field sits in \mathbb{C} in three ways.

Example 2

A cube root and its inverse

Let \displaystyle a = \sqrt[3]{7}. Find [\mathbb{Q}(a) : \mathbb{Q}] and a basis, write (1 + a)^{-1} in it, and show that \displaystyle \mathbb{Q}(\sqrt[3]{49}) = \mathbb{Q}(a).

Solution

At \pm 1, \pm 7, x^3 - 7 takes -6, -8, 336, -350, so it has no rational root and is irreducible. By Theorem 5 the degree is 3, with basis \displaystyle (1, \sqrt[3]{7}, \sqrt[3]{49}).

The sum of cubes gives (1 + a)(1 - a + a^2) = 1 + a^3 = 8, so

(1 + a)^{-1} = \tfrac{1}{8} - \tfrac{1}{8}\sqrt[3]{7} + \tfrac{1}{8}\sqrt[3]{49}.

Since \displaystyle \sqrt[3]{49} = a^2, \mathbb{Q}(a^2) \subseteq \mathbb{Q}(a); and a = a^4/7 = (a^2)^2/7 \in \mathbb{Q}(a^2). (Corollary 7(c) also gives it: the degree 3 is prime and a^2 \notin \mathbb{Q}.)

Example 3

Coprime degrees multiply

Find \displaystyle [\mathbb{Q}(\sqrt{3}, \sqrt[3]{7}) : \mathbb{Q}] and a basis. Deduce that \displaystyle \sqrt{3} \notin \mathbb{Q}(\sqrt[3]{7}), and write \displaystyle (\sqrt{3} + \sqrt[3]{7})^2 in the basis.

Solution

x^2 - 3 and x^3 - 7 are irreducible (no rational roots), of coprime degrees 2 and 3, so Corollary 8 gives degree 6. In the tower \displaystyle \mathbb{Q} \subseteq \mathbb{Q}(\sqrt[3]{7}) \subseteq E = \mathbb{Q}(\sqrt{3}, \sqrt[3]{7}), Theorem 6 gives \displaystyle [E : \mathbb{Q}(\sqrt[3]{7})] = 6/3 = 2. Were \displaystyle \sqrt{3} in \displaystyle \mathbb{Q}(\sqrt[3]{7}), Lemma 2 would give \displaystyle E = \mathbb{Q}(\sqrt[3]{7}), of degree 1 over it. So x^2 - 3 has no root in \displaystyle \mathbb{Q}(\sqrt[3]{7}), is irreducible there, and \displaystyle (1, \sqrt{3}) is a basis of E over it. Theorem 6 gives the basis

\big(1, \sqrt[3]{7}, \sqrt[3]{49}, \sqrt{3}, \sqrt{3}\sqrt[3]{7}, \sqrt{3}\sqrt[3]{49}\big).

Expanding, \displaystyle (\sqrt{3} + \sqrt[3]{7})^2 = 3 + \sqrt[3]{49} + 2\sqrt{3}\sqrt[3]{7}, with coordinates (3, 0, 1, 0, 2, 0). The figure above shows both routes up: 2 \cdot 3 = 3 \cdot 2.

Example 4

One generator for two square roots

Let \displaystyle E = \mathbb{Q}(\sqrt{5}, i) and \displaystyle \gamma = \sqrt{5} + i. Find [E : \mathbb{Q}], show that E = \mathbb{Q}(\gamma), and show that x^4 - 8x^2 + 36 is irreducible over \mathbb{Q}.

Solution

x^2 - 5 has no rational root, so \displaystyle [\mathbb{Q}(\sqrt{5}) : \mathbb{Q}] = 2. As \displaystyle \mathbb{Q}(\sqrt{5}) \subseteq \mathbb{R}, x^2 + 1 has no root there, so \displaystyle [E : \mathbb{Q}(\sqrt{5})] = 2 and [E : \mathbb{Q}] = 4 (Theorem 6), with basis \displaystyle (1, \sqrt{5}, i, \sqrt{5}\,i).

Clearly \mathbb{Q}(\gamma) \subseteq E. Now \displaystyle \gamma^2 = 4 + 2\sqrt{5}\,i and \displaystyle \gamma^3 = 2\sqrt{5} + 14i, so

i = \frac{\gamma^3 - 2\gamma}{12}, \qquad \sqrt{5} = \gamma - i = \frac{14\gamma - \gamma^3}{12},

and E = \mathbb{Q}(\gamma). From \displaystyle \gamma^2 - 4 = 2\sqrt{5}\,i, (\gamma^2 - 4)^2 = -20: \gamma is a root of p = x^4 - 8x^2 + 36. It is a root of some irreducible factor h of p, and Theorem 5 gives \deg h = [\mathbb{Q}(\gamma) : \mathbb{Q}] = 4, so p is a constant times h, irreducible.

The roots \displaystyle \sqrt{5}, i, \displaystyle \sqrt{5}\,i of x^2 - 5, x^2 + 1, x^2 + 5 give three intermediate fields of degree 2, each with E of degree 4/2 = 2 over it:

2
2
2
2
2
2
\displaystyle \mathbb{Q}(\sqrt{5}, i)
\displaystyle \mathbb{Q}(\sqrt{5})
\mathbb{Q}(i)
\displaystyle \mathbb{Q}(i\sqrt{5})
\mathbb{Q}
Figure 2The field \displaystyle \mathbb{Q}(\sqrt{5}, i) = \mathbb{Q}(\sqrt{5} + i) of Example 4 and three of its subfields. Every edge has degree 2, and every path from bottom to top multiplies to 4.

That there are no others is a theorem of Galois theory (The fundamental theorem of Galois theory), not of degrees alone.

Example 5

Every quadratic extension adjoins a square root

Let \operatorname{char} F \ne 2 and [E : F] = 2. Prove that E = F(\delta) for some \delta with \delta^2 \in F. Find such a \delta when E = \mathbb{Q}(\alpha) with \alpha^2 + 3\alpha + 1 = 0.

Solution

Proof. Pick \alpha \in E \setminus F. The list (1, \alpha, \alpha^2) is longer than 2, so dependent (Result 5 of Bases and dimension, proved); as (1, \alpha) is independent, the relation involves \alpha^2, and \alpha^2 + b\alpha + c = 0 with b, c \in F. Completing the square as in Deriving the quadratic formula, put \delta = 2\alpha + b. Then

\delta^2 = 4(\alpha^2 + b\alpha + c) + b^2 - 4c = b^2 - 4c \in F.

Since 2 \ne 0, \alpha = (\delta - b)/2 \in F(\delta), so \delta \notin F, and F(\delta) = E by Corollary 7(c). \square

For \alpha^2 + 3\alpha + 1 = 0: \delta = 2\alpha + 3 and \delta^2 = 9 - 4 = 5, so \displaystyle \mathbb{Q}(\alpha) = \mathbb{Q}(\sqrt{5}).

Example 6

Square roots never reach a cube root

Let F = K_0 \subseteq K_1 \subseteq \dots \subseteq K_r with every [K_i : K_{i-1}] \le 2. Prove that no root of an irreducible cubic in F[x] lies in K_r. Deduce it for x^3 - 4x + 2 over \mathbb{Q}, although all three of its roots are real.

Solution

Proof. By Theorem 6 and Proof by induction, [K_r : F] = \prod_i [K_i : K_{i-1}], a product of 1’s and 2’s, so [K_r : F] = 2^k. A root \theta \in K_r of an irreducible cubic would give 3 \mid 2^k by Corollary 7(b), which is false. \square

For g = x^3 - 4x + 2, the candidates \pm 1, \pm 2 give -1, 5, 2, 2, so g is irreducible. Since g(-3) = -13 < 0 < g(0) = 2 and g(1) = -1 < 0 < g(2) = 2, all three roots are real; even towers inside \mathbb{R} miss them. This is the shape of the argument in Compass-and-straightedge constructions, and why an angle cannot be trisected.

Boundaries

  • Irreducibility is needed in Theorems 3 and 5. \displaystyle \sqrt{3} is a root of x^4 - 9, yet \displaystyle [\mathbb{Q}(\sqrt{3}) : \mathbb{Q}] = 2; and \mathbb{Q}[x]/\langle x^4 - 9 \rangle is not a field, as x^2 - 3 and x^2 + 3 give zero divisors.
  • Infinite degree. [\mathbb{R} : \mathbb{Q}] is infinite: coordinates would put \mathbb{R} in bijection with some \mathbb{Q}^n, which is countable (Countable sets). For rational functions, [F(t) : F] is infinite and F[t] \ne F(t) (Exercise 13): Theorem 5 needs \alpha to be a root of some nonzero polynomial, the subject of Algebraic elements and minimal polynomials.
  • Degrees of generators need not multiply. \displaystyle \mathbb{Q}(\sqrt{2}, \sqrt{8}) = \mathbb{Q}(\sqrt{2}) has degree 2; without coprimality Corollary 8 gives only the bound and the divisibility (Exercise 6).
  • Characteristic 2. Example 5 divides by 2; over \mathbb{Z}_2 it fails (Exercise 14).
  • Which root. Different roots of one f can generate different subfields, all isomorphic to Kronecker’s field (Example 1); comparing them is the work of Splitting fields and Field automorphisms and the Galois group.

Where this goes next

Algebraic elements and minimal polynomials shows that every element of a finite extension is a root of an irreducible polynomial, its minimal polynomial, so Theorem 5 measures every simple subextension. Splitting fields repeats Corollary 4 until a polynomial factors completely, and Finite fields uses Theorem 1(c) to count elements. Example 6 is the core of Compass-and-straightedge constructions, and why an angle cannot be trisected, and lattices like that of Example 4 are what the The fundamental theorem of Galois theory matches with subgroups.

Exercises

Computing in F[x]/\langle f \rangle

  1. Show that f = x^3 + x + 3 is irreducible over \mathbb{Q}. In K = \mathbb{Q}[x]/\langle f \rangle with \alpha = x + \langle f \rangle, write \alpha^4 and \alpha^5 in the basis (1, \alpha, \alpha^2).
  2. In the field K of Exercise 1, find \alpha^{-1} and (\alpha - 1)^{-1} in the same basis.
  3. Show that K = \mathbb{Z}_2[x]/\langle x^3 + x^2 + 1 \rangle is a field with 8 elements. With \beta = x + \langle x^3 + x^2 + 1 \rangle, write \beta^3, \dots, \beta^7 in the basis (1, \beta, \beta^2) and find \beta^{-1}.
  4. Writing k for [k], show that K = \mathbb{Z}_7[x]/\langle x^2 + x + 3 \rangle is a field with 49 elements, and with \gamma = x + \langle x^2 + x + 3 \rangle find \gamma^{-1} and \gamma^3 in the basis (1, \gamma).

Degrees and bases

  1. Find \displaystyle [\mathbb{Q}(\sqrt[3]{11}) : \mathbb{Q}] and a basis, and write \displaystyle (\sqrt[3]{11} - 2)^{-1} in it.
  2. At a hackathon, Naomi and Kofi write a program for degrees of fields. For \displaystyle [\mathbb{Q}(\sqrt{6}, \sqrt{10}, \sqrt{15}) : \mathbb{Q}], Kofi’s version multiplies the three degrees and prints 8; Naomi’s prints 4. Which is right? Prove it.
  3. Let \displaystyle \beta = \sqrt{1 + \sqrt{3}}. Show that [\mathbb{Q}(\beta) : \mathbb{Q}] = 4 and that x^4 - 2x^2 - 2 is irreducible over \mathbb{Q}.
  4. Let \alpha \in \mathbb{C} be a root of x^3 - x - 1. Find \displaystyle [\mathbb{Q}(\alpha, \sqrt{2}) : \mathbb{Q}] and a basis. Can a subfield of \mathbb{C} of degree 4 or 10 over \mathbb{Q} contain a root of x^3 - x - 1?
  5. Let K = \mathbb{Q}[x]/\langle x^2 - 2x - 2 \rangle with \alpha = x + \langle x^2 - 2x - 2 \rangle. Find \delta = 2\alpha + b with \delta^2 \in \mathbb{Q}, and \eta \in K with \eta^2 = 3.
  6. Show that x^3 - 5 is irreducible over \displaystyle \mathbb{Q}(\sqrt{2}, \sqrt{3}), while x^2 - 6 is not.

Proofs

  1. Let F \subseteq K \subseteq E with E/F finite. Prove that [K : F] = [E : F] forces K = E, and [E : K] = [E : F] forces K = F.
  2. Let E/F be an extension and \alpha \in E with [F(\alpha) : F] finite and odd. Prove that F(\alpha^2) = F(\alpha).
  3. Let F(t) = \operatorname{Frac}(F[t]) be the field of rational functions in an indeterminate t. Prove that [F(t) : F] is infinite, that 1/t \notin F[t], and that t is a root of no nonzero polynomial in F[x].
  4. Let K = \mathbb{Z}_2[x]/\langle x^2 + x + 1 \rangle, with \beta = x + \langle x^2 + x + 1 \rangle. Prove that [K : \mathbb{Z}_2] = 2 but no \delta \in K \setminus \mathbb{Z}_2 has \delta^2 \in \mathbb{Z}_2, and say where the proof of Example 5 breaks.
  5. Let [E : F] = n and let f \in F[x] be irreducible of degree m with \gcd(m, n) = 1. Prove that f is irreducible in E[x].
Answers
  1. At \pm 1, \pm 3, f takes 5, 1, 33, -27, so the cubic has no rational root and is irreducible. From \alpha^3 = -\alpha - 3: \alpha^4 = -\alpha^2 - 3\alpha and \alpha^5 = -\alpha^3 - 3\alpha^2 = -3\alpha^2 + \alpha + 3.
  2. \alpha(\alpha^2 + 1) = \alpha^3 + \alpha = -3, so \displaystyle \alpha^{-1} = -\frac{1}{3}\alpha^2 - \frac{1}{3}. Division gives f = (x - 1)(x^2 + x + 2) + 5, so (\alpha - 1)(\alpha^2 + \alpha + 2) = -5 and \displaystyle (\alpha - 1)^{-1} = -\frac{1}{5}\alpha^2 - \frac{1}{5}\alpha - \frac{2}{5}.
  3. The cubic is 1 at 0 and at 1, so irreducible, and K is a field with 2^3 = 8 elements (Theorems 3 and 1(c)). From \beta^3 = \beta^2 + 1: \beta^4 = \beta^2 + \beta + 1, \beta^5 = \beta + 1, \beta^6 = \beta^2 + \beta, \beta^7 = 1. So \beta^{-1} = \beta^6 = \beta^2 + \beta, and the seven powers are the seven nonzero elements.
  4. The values at 0, \dots, 6 are 3, 5, 2, 1, 2, 5, 3, never 0, so the quadratic is irreducible and K has 7^2 = 49 elements. As \gamma^2 = 6\gamma + 4, \gamma(\gamma + 1) = -3 = 4 and 4 \cdot 2 = 1, so \gamma^{-1} = 2\gamma + 2. And \gamma^3 = 6\gamma^2 + 4\gamma = 40\gamma + 24 = 5\gamma + 3.
  5. x^3 - 11 has no rational root (\pm 1, \pm 11 fail): degree 3, basis \displaystyle (1, \sqrt[3]{11}, \sqrt[3]{121}). As \displaystyle (\sqrt[3]{11} - 2)(\sqrt[3]{121} + 2\sqrt[3]{11} + 4) = 11 - 8 = 3, the inverse is \displaystyle \frac{4}{3} + \frac{2}{3}\sqrt[3]{11} + \frac{1}{3}\sqrt[3]{121}.
  6. Naomi: the degree is 4. As \displaystyle \sqrt{15} = \sqrt{6}\sqrt{10}/2, the field is \displaystyle \mathbb{Q}(\sqrt{6}, \sqrt{10}). If \displaystyle \sqrt{10} = a + b\sqrt{6} with a, b \in \mathbb{Q}, then \displaystyle 10 = a^2 + 6b^2 + 2ab\sqrt{6}, so ab = 0; b = 0 makes \displaystyle \sqrt{10} rational, and a = 0 gives b^2 = 5/3, not a rational square. So \displaystyle [\mathbb{Q}(\sqrt{6}, \sqrt{10}) : \mathbb{Q}(\sqrt{6})] = 2, and the degree is 2 \cdot 2 = 4.
  7. \displaystyle \beta^2 = 1 + \sqrt{3}, so \displaystyle \mathbb{Q}(\sqrt{3}) \subseteq \mathbb{Q}(\beta). If \displaystyle \beta = a + b\sqrt{3} with a, b rational, then 2ab = 1 and a^2 + 3b^2 = 1, so 12b^4 - 4b^2 + 1 = 0, which has no real b^2 (16 - 48 < 0). So \displaystyle [\mathbb{Q}(\beta) : \mathbb{Q}(\sqrt{3})] = 2 and the degree is 4. As (\beta^2 - 1)^2 = 3, \beta is a root of x^4 - 2x^2 - 2, irreducible as in Example 4.
  8. At \pm 1 the cubic takes -1, -1, so it is irreducible; with x^2 - 2 and \gcd(3, 2) = 1, Corollary 8 gives 6, with basis \displaystyle (1, \alpha, \alpha^2, \sqrt{2}, \sqrt{2}\alpha, \sqrt{2}\alpha^2) (Theorem 6). No: by Corollary 7(b), 3 would divide 4 or 10.
  9. \delta = 2\alpha - 2: \delta^2 = 4(\alpha^2 - 2\alpha) + 4 = 4 \cdot 2 + 4 = 12. Then \eta = \delta/2 = \alpha - 1.
  10. \displaystyle \mathbb{Q} \subseteq \mathbb{Q}(\sqrt{2}) \subseteq \mathbb{Q}(\sqrt{2}, \sqrt{3}) has steps of degree at most 2, and x^3 - 5 is irreducible over \mathbb{Q} (\pm 1, \pm 5 are not roots), so by Example 6 it has no root in the top field, hence is irreducible there. But \displaystyle x^2 - 6 = (x - \sqrt{2}\sqrt{3})(x + \sqrt{2}\sqrt{3}).
  11. Proof. By Theorem 6, [E : F] = [E : K][K : F], all finite. If [K : F] = [E : F], then [E : K] = 1 and K = E (Theorem 1(b)). If [E : K] = [E : F], then [K : F] = 1 and K = F. \square
  12. Proof. F \subseteq F(\alpha^2) \subseteq F(\alpha), and \alpha is a root of x^2 - \alpha^2 \in F(\alpha^2)[x], so of an irreducible factor of degree 1 or 2; by Theorem 5 and Lemma 2, [F(\alpha) : F(\alpha^2)] \le 2. It divides the odd [F(\alpha) : F] (Corollary 7(a)), so it is 1 and F(\alpha^2) = F(\alpha). \square
  13. Proof. c_0 + c_1t + \dots + c_nt^n with c_k \in F is the polynomial with those coefficients, 0 only if every c_k = 0. So (1, t, \dots, t^n) is independent for every n, while a space of finite dimension d has no independent list longer than d (Result 5): [F(t) : F] is infinite. If 1/t = g \in F[t], then tg = 1, but tg is 0 or of degree at least 1. A nonzero g \in F[x] gives g(t) = g \ne 0. \square
  14. Proof. x^2 + x + 1 is 1 at 0 and 1, so irreducible: [K : \mathbb{Z}_2] = 2 and K = \{0, 1, \beta, \beta + 1\}. Now \beta^2 = \beta + 1 and (\beta + 1)^2 = \beta^2 + 2\beta + 1 = \beta, neither in \mathbb{Z}_2. In Example 5, 2 = 0 makes \delta = 2\alpha + b = b \in F, and (\delta - b)/2 is undefined. \square
  15. Proof. Let p be an irreducible factor of f in E[x] and L an extension of E with a root \theta of p (Corollary 4); f(\theta) = 0. By Theorem 5, [E(\theta) : E] = \deg p and [F(\theta) : F] = m. Theorem 6 on F \subseteq E \subseteq E(\theta) and on F \subseteq F(\theta) \subseteq E(\theta) makes [E(\theta) : F] = n \deg p a multiple of m. As \gcd(m, n) = 1, m \mid \deg p \le m, so \deg p = m and f is a constant times p: irreducible in E[x]. \square
math.wiki · updated 2026-11-17 · CC-BY-SA-4.0