math.wiki
Abstract algebra › Galois theory › Why the quintic has no formula
Sign in

Why the quintic has no formula

The quadratic formula solves every quadratic with one square root. Is there a formula of the same kind, built from the coefficients with sums, products, quotients and roots, that solves every equation of degree five?

Where you are 4 in Galois theory

Why this exists

Completing the square writes the roots of x^2 + bx + c as \displaystyle \frac{-b \pm \sqrt{b^2 - 4c}}{2}: add, multiply, divide and take one square root. Formulas of the same kind were published for the cubic and the quartic in the sixteenth century. For degree five the search failed, and for a reason: Abel proved in the 1820s that no such formula exists, and Galois found the criterion in terms of groups that this lesson proves. A formula in radicals puts the roots in a tower of fields, each step adjoining an n-th root; such a tower forces the Galois group to be solvable; and a quintic as plain as x^5 - 4x^3 + 2 has Galois group S_5, which is not.

Definition

In plain words: a field is reached by radicals when it can be built from the base field by adjoining, one at a time, numbers some power of which already lies in the field built so far; a polynomial is solvable by radicals when all its roots lie in such a field.

Throughout, every field is a subfield of \mathbb{C}. It contains \mathbb{Q}, so every irreducible polynomial over it is separable (Field automorphisms and the Galois group, Theorem 6), and the splitting field of f \in F[x] is taken inside \mathbb{C}, where it is F(\alpha_1, \dots, \alpha_n) for the complex roots \alpha_i of f (The fundamental theorem of algebra; Splitting fields, Lemma 2(b)). For a positive integer n, \displaystyle \omega_n = \operatorname{cis}\frac{2\pi}{n} and \mu_n = \{z \in \mathbb{C} : z^n = 1\} = \{1, \omega_n, \dots, \omega_n^{n-1}\}, a cyclic group of order n (Roots of a complex number and roots of unity).

  • An extension L/F is radical if L = F(\alpha_1, \dots, \alpha_m) for some \alpha_i \in L and positive integers n_i with \alpha_i^{n_i} \in F(\alpha_1, \dots, \alpha_{i-1}) for each i (for i = 1, the field F).
  • A nonconstant f \in F[x] is solvable by radicals over F if its splitting field E lies in some radical extension L of F.
  • The Galois group of f over F is \operatorname{Gal}(E/F).

Every quadratic is solvable by radicals: the roots of x^2 + bx + c lie in \displaystyle F(\sqrt{b^2 - 4c}), one step with n_1 = 2.

Intuition

One radical step K \subseteq K(\beta) with \beta^n \in K has a small group once K contains \omega_n: an automorphism can only multiply \beta by an n-th root of unity, so the group sits inside the cyclic group \mu_n (Theorem 2). Adjoining roots of unity gives an abelian group (Theorem 1). A radical tower, enlarged so that it is Galois with its roots of unity first (Lemma 3), is a chain of abelian steps, and the fundamental theorem turns it into a subnormal series with abelian factors (Theorem 4).

r_1
r_2
r_3
z
\bar z
Re Im
Figure 1Complex conjugation fixes the three real roots of x^5 - 4x^3 + 2 and swaps z with \bar z: a transposition in the Galois group.

On the other side, the Galois group of an irreducible quintic contains a 5-cycle, and when exactly two roots are non-real, complex conjugation swaps them and fixes the other three: a transposition. Together these generate S_5 (Theorem 5), which is not solvable.

Results

Every field is a subfield of \mathbb{C}.

Theorem 1 (roots of unity). Let F be a field, n a positive integer and \omega = \omega_n. Then F(\omega) is a splitting field of x^n - 1 over F. For each \sigma \in \operatorname{Gal}(F(\omega)/F) there is an integer k with \sigma(\omega) = \omega^k, and \sigma \mapsto [k] is a well-defined injective homomorphism \operatorname{Gal}(F(\omega)/F) \to \mathbb{Z}_n^\times. In particular \operatorname{Gal}(F(\omega)/F) is abelian.

Theorem 2 (pure equations). Let F be a field containing \omega_n, let a \in F be nonzero and let \beta \in \mathbb{C} satisfy \beta^n = a. Then F(\beta) is a splitting field of x^n - a over F, and \sigma \mapsto \sigma(\beta)/\beta is an injective homomorphism \operatorname{Gal}(F(\beta)/F) \to \mu_n. In particular \operatorname{Gal}(F(\beta)/F) is cyclic of order dividing n.

Lemma 3 (a Galois radical tower). Let L/F be radical, with \alpha_i and n_i as in the definition, and let N = n_1 n_2 \cdots n_m. There is a field M \supseteq L with M/F Galois, and elements \beta_1, \dots, \beta_r of M, such that M = F(\omega_N, \beta_1, \dots, \beta_r) and, for each j, \beta_j^{d_j} \in F(\omega_N, \beta_1, \dots, \beta_{j-1}) for some positive d_j dividing N.

Theorem 4 (radicals give solvable groups). If f \in F[x] is solvable by radicals over F, then the Galois group of f over F is solvable.

Theorem 5 (generating S_p). Let p be prime. A subgroup of S_p that contains a p-cycle and a transposition is S_p.

Theorem 6 (the criterion). Let p be prime and f \in \mathbb{Q}[x] irreducible of degree p with exactly two non-real roots in \mathbb{C}. Then the Galois group of f over \mathbb{Q} is isomorphic to S_p.

Corollary 7 (the quintic). (a) An irreducible f \in \mathbb{Q}[x] of degree 5 with exactly three real roots is not solvable by radicals over \mathbb{Q}. (b) No expression built from a_0, \dots, a_4 and rational numbers by addition, subtraction, multiplication, division and n-th roots has, among its values over all choices of its n-th roots, the five roots of every monic quintic x^5 + a_4x^4 + \dots + a_0 with rational coefficients.

Why it is true

Proof of Theorem 1. The roots of x^n - 1 are 1, \omega, \dots, \omega^{n-1}; they lie in F(\omega) and generate it, so F(\omega) is a splitting field. Let \sigma \in \operatorname{Gal}(F(\omega)/F). As \sigma(\omega)^n = \sigma(\omega^n) = 1, \sigma(\omega) = \omega^k for some k, determined modulo n. Restricted to \mu_n, \sigma is an injective homomorphism of multiplicative groups, so \omega^k also has order n; as \operatorname{ord}(\omega^k) = n/\gcd(k, n) (Cyclic groups), [k] \in \mathbb{Z}_n^\times. If \sigma(\omega) = \omega^k and \tau(\omega) = \omega^l, then \sigma\tau(\omega) = \sigma(\omega)^l = \omega^{kl}: a homomorphism. Every element of F(\omega) is a polynomial in \omega with coefficients in F, which \sigma fixes, so \sigma is determined by \sigma(\omega) and the kernel is trivial. A group isomorphic to a subgroup of the abelian group \mathbb{Z}_n^\times is abelian. \square

Proof of Theorem 2. Write \omega = \omega_n. Since \beta \ne 0, the \beta\omega^k with 0 \le k < n are n distinct roots of x^n - a, hence all of them (Polynomial rings and the division algorithm, Theorem 6). They lie in F(\beta), as \omega \in F, and generate it, so F(\beta) is a splitting field. For \sigma in the group, \sigma(\beta) is a root, so \sigma(\beta)/\beta \in \mu_n. If \sigma(\beta) = \beta\zeta and \tau(\beta) = \beta\eta with \zeta, \eta \in \mu_n \subseteq F, then \sigma\tau(\beta) = \sigma(\beta)\sigma(\eta) = \beta\zeta\eta, because \sigma fixes \eta: a homomorphism. It is injective because \sigma is determined by \sigma(\beta). Its image is a subgroup of the cyclic group \mu_n, so it is cyclic of order dividing n (Cyclic groups, Lagrange's theorem). \square

Proof of Lemma 3. Let g be the product of the minimal polynomials m_{\alpha_i}(x) over F, and M the splitting field of (x^N - 1)g(x) over F. Then M/F is Galois (definition in The fundamental theorem of Galois theory) and M \supseteq L. Let G = \operatorname{Gal}(M/F). By The fundamental theorem of Galois theory, Theorem 4, the roots of m_{\alpha_i} are exactly the \sigma(\alpha_i) with \sigma \in G, and the roots of x^N - 1 are the powers of \omega_N, so M is generated over F by \omega_N and the \sigma(\alpha_i). After \omega_N, adjoin all \sigma(\alpha_1), then all \sigma(\alpha_2), and so on, calling them \beta_1, \dots, \beta_r in that order. If \beta_j = \sigma(\alpha_i), take d_j = n_i, which divides N. Now \alpha_i^{n_i} is a rational expression, with coefficients in F, in \alpha_1, \dots, \alpha_{i-1}; applying \sigma, which fixes F, \beta_j^{d_j} = \sigma(\alpha_i^{n_i}) is the same expression in \sigma(\alpha_1), \dots, \sigma(\alpha_{i-1}), all adjoined earlier. \square

Proof of Theorem 4. The splitting field E of f lies in a radical extension L of F; take M, N, \beta_j and d_j as in Lemma 3. Set M_0 = F, M_1 = F(\omega_N) and M_{j+1} = M_j(\beta_j), so that M_{r+1} = M, and let H_j = \operatorname{Gal}(M/M_j). Then

G = H_0 \ge H_1 \ge \dots \ge H_{r+1} = \{e\}.

Each M/M_j is Galois, as M is a splitting field over M_j (Splitting fields, Lemma 2(c)). Each step M_{j+1}/M_j is a splitting field with abelian group: for j = 0 by Theorem 1; for j \ge 1, if \beta_j = 0 the step is trivial, and otherwise \omega_{d_j} = \omega_N^{N/d_j} \in M_j, so Theorem 2 applies with a = \beta_j^{d_j}. So M_{j+1}/M_j is normal (Splitting fields, Theorem 6), and The fundamental theorem of Galois theory, Theorem 5(d), for the Galois extension M/M_j with group H_j, gives H_{j+1} \trianglelefteq H_j and H_j/H_{j+1} \cong \operatorname{Gal}(M_{j+1}/M_j), abelian. Hence G is solvable. Finally E/F is normal, so the same theorem gives \operatorname{Gal}(E/F) \cong G/\operatorname{Gal}(M/E), solvable by Solvable groups, Theorem 4(b). \square

Proof of Theorem 5. Let \sigma be the p-cycle and \tau = (a\ b) the transposition in the subgroup G. The p-cycle carries a through all p points, so \sigma^k(a) = b for some 1 \le k \le p - 1. As p is prime, \rho = \sigma^k has order p; the order of a permutation is the least common multiple of its cycle lengths (Permutation groups and cycle notation), so \rho is a p-cycle, \rho = (a\ b\ c_3\ \cdots\ c_p). Let \pi \in S_p send a, b, c_3, \dots, c_p to 1, 2, 3, \dots, p. Conjugation relabels cycles (Normal subgroups, Example 4), so \pi G\pi^{-1} contains (1\ 2\ \cdots\ p) and (1\ 2). Conjugating (1\ 2) by powers of (1\ 2\ \cdots\ p) gives (2\ 3), \dots, (p{-}1\ \ p), which generate S_p by the argument of Solvable groups, Answer 7, valid for any number of points. So \pi G\pi^{-1} = S_p and G = S_p. \square

Proof of Theorem 6. Let E be the splitting field and G = \operatorname{Gal}(E/\mathbb{Q}). The p roots are distinct (Field automorphisms and the Galois group, Theorem 6), and restriction to the set R of roots is an injective homomorphism G \to \operatorname{Sym}(R) (Theorem 2(b) there). By Field automorphisms and the Galois group, Exercise 11, some element of G acts on R as a p-cycle. Complex conjugation is an automorphism of \mathbb{C} fixing \mathbb{Q}, and it sends roots of f to roots, as f has real coefficients (Complex roots come in pairs); so it maps E, which R generates, into E, and its restriction lies in G (Field automorphisms and the Galois group, Theorem 1(d)). It fixes the p - 2 real roots and swaps the two non-real ones: a transposition of R. Numbering the roots identifies \operatorname{Sym}(R) with S_p, and Theorem 5 shows that the image of G is all of it. \square

Proof of Corollary 7. (a) Non-real roots come in conjugate pairs, so three real roots leave exactly two non-real ones. By Theorem 6 the group is isomorphic to S_5, which is not solvable (Solvable groups, Corollary 9), so Theorem 4 rules out solvability by radicals. (b) Suppose such an expression existed, and evaluate it at the coefficients of f = x^5 - 4x^3 + 2, which satisfies (a) by Example 3. For one choice of the n-th roots, the computation adjoins, one at a time, a \beta with \beta^n in the field built so far; so the final value lies in a radical extension of \mathbb{Q}. Each root \alpha_k of f is such a value, so \alpha_k \in L_k with L_k/\mathbb{Q} radical. Adjoining the radicals of L_1, then those of L_2, and so on, gives one radical extension of \mathbb{Q} (each power lies in a smaller field built earlier) containing all five roots, hence the splitting field, against (a). \square

Worked examples

Example 1

Twelfth roots of unity

Let \omega = \omega_{12}. Find \operatorname{Gal}(\mathbb{Q}(\omega)/\mathbb{Q}) and the fixed field of each element.

Solution

\displaystyle \omega = \frac{\sqrt{3}}{2} + \frac{1}{2}i, and conversely i = \omega^3 and \displaystyle \sqrt{3} = \omega + \omega^{-1}, so \displaystyle \mathbb{Q}(\omega) = \mathbb{Q}(\sqrt{3}, i), of degree 4 by the tower law, since \displaystyle \mathbb{Q}(\sqrt{3}) \subseteq \mathbb{R} has degree 2 and i \notin \mathbb{R}. The group has order 4 (Field automorphisms and the Galois group, Corollary 7), and Theorem 1 embeds it in \mathbb{Z}_{12}^\times = \{1, 5, 7, 11\}, also of order 4; so each k there gives exactly one \sigma_k with \sigma_k(\omega) = \omega^k. As 5^2, 7^2 and 11^2 are all \equiv 1 \pmod{12}, the group is V_4: abelian, as Theorem 1 says, but not cyclic.

Using \omega^{12} = 1: \sigma_5(i) = \omega^{15} = \omega^3 = i and \sigma_7(i) = \omega^{21} = \omega^9 = -i, while \displaystyle \sigma_5(\sqrt{3}) = 2\cos\frac{5\pi}{6} = -\sqrt{3} and \displaystyle \sigma_7(\sqrt{3}) = 2\cos\frac{7\pi}{6} = -\sqrt{3}; and \sigma_{11}(\omega) = \omega^{-1} = \bar\omega, so \sigma_{11} is complex conjugation. A subgroup of order 2 fixes a field of degree 2 over \mathbb{Q} (The fundamental theorem of Galois theory, Theorem 5(b)), so the fixed fields are \mathbb{Q}(i) for \sigma_5, \displaystyle \mathbb{Q}(\sqrt{3}\,i) for \sigma_7, and \displaystyle \mathbb{Q}(\sqrt{3}) for \sigma_{11}.

Example 2

A pure quartic

Let \displaystyle \beta = \sqrt[4]{6}, the positive real root, and E = \mathbb{Q}(i, \beta). Show that E is the splitting field of x^4 - 6 over \mathbb{Q}, find \operatorname{Gal}(E/\mathbb{Q}(i)), and give a subnormal series of G = \operatorname{Gal}(E/\mathbb{Q}) with abelian factors.

Solution

Since i = \omega_4, the roots are \pm\beta, \pm i\beta, and they generate E. x^4 - 6 is Eisenstein at 3 (Irreducible polynomials and Eisenstein's criterion), so [\mathbb{Q}(\beta) : \mathbb{Q}] = 4; \mathbb{Q}(\beta) \subseteq \mathbb{R} does not contain i, so [E : \mathbb{Q}] = 8 and [E : \mathbb{Q}(i)] = 4. Theorem 2, with F = \mathbb{Q}(i), embeds H = \operatorname{Gal}(E/\mathbb{Q}(i)) in \mu_4 = \{\pm 1, \pm i\}, and |H| = 4, so H is cyclic, generated by \tau with \tau(\beta) = i\beta and \tau(i) = i.

\mathbb{Q}(i) is the splitting field of x^2 + 1, so H \trianglelefteq G and G/H \cong \operatorname{Gal}(\mathbb{Q}(i)/\mathbb{Q}), of order 2 (The fundamental theorem of Galois theory, Theorem 5(d)). The series G \ge H \ge \{e\}, matching the radical tower \mathbb{Q} \subseteq \mathbb{Q}(i) \subseteq E, has cyclic factors of orders 2 and 4. Complex conjugation c fixes \beta and sends i to -i, so c\tau c(\beta) = -i\beta and c\tau c = \tau^{-1}: with r = \tau and s = c, G is a copy of D_4.

Example 3

A quintic with Galois group S_5

Show that f = x^5 - 4x^3 + 2 has Galois group S_5 over \mathbb{Q}, and so is not solvable by radicals.

Solution

f is Eisenstein at 2, so it is irreducible. Its derivative f' = x^2(5x^2 - 12) vanishes at 0 and \pm c with \displaystyle c = \sqrt{12/5}, is positive for |x| > c, and is negative on (-c, c) except at 0. So f is strictly increasing on (-\infty, -c], strictly decreasing on [-c, c] and strictly increasing on [c, \infty). With \displaystyle c^2 = \frac{12}{5},

f(\pm c) = \pm c^3(c^2 - 4) + 2 = 2 \mp \tfrac{8}{5}c^3,

and \displaystyle c^3 = c \cdot c^2 > \frac{12}{5} because c > 1, so \displaystyle f(c) < 2 - \frac{96}{25} < 0 < f(-c). As f \to -\infty at the left and +\infty at the right, f has exactly one root on each interval: three real roots, near -2.057, 0.848 and 1.929, and two non-real ones, near -0.360 \pm 0.682i. Theorem 6 gives S_5, and Corollary 7(a) says f is not solvable by radicals.

r_1
r_2
r_3
x y
Figure 2Example 3: f rises, falls, then rises, with f(-c) > 0 > f(c) at \displaystyle c = \sqrt{12/5}, so it has exactly three real roots.
Example 4

A quintic that is solvable

Let \displaystyle \beta = \sqrt[5]{7}, the real root, and \zeta = \omega_5. Find the degree over \mathbb{Q} of the splitting field E of x^5 - 7, and show that its Galois group G is solvable.

Solution

The roots are \beta\zeta^k, 0 \le k \le 4, so E = \mathbb{Q}(\beta, \zeta). x^5 - 7 is Eisenstein at 7, so [\mathbb{Q}(\beta) : \mathbb{Q}] = 5; \zeta is a root of x^4 + x^3 + x^2 + x + 1, irreducible (Irreducible polynomials and Eisenstein's criterion, Theorem 6), so [\mathbb{Q}(\zeta) : \mathbb{Q}] = 4. Both divide [E : \mathbb{Q}], so 20 does; and [E : \mathbb{Q}(\zeta)] \le 5. So [E : \mathbb{Q}] = 20 = |G|.

Over K = \mathbb{Q}(\zeta), Theorem 2 makes H = \operatorname{Gal}(E/K) cyclic of order 5. K is a splitting field (Theorem 1), so H \trianglelefteq G and G/H \cong \operatorname{Gal}(K/\mathbb{Q}), which Theorem 1 embeds in \mathbb{Z}_5^\times; both have order 4, and \mathbb{Z}_5^\times is cyclic, generated by 2 (powers 2, 4, 3, 1). So G \ge H \ge \{e\} has cyclic factors of orders 4 and 5: degree five alone is no obstruction. Theorem 6 is silent here, since x^5 - 7 is increasing, with one real root and four non-real ones.

Example 5

Infinitely many quintics with group S_5

Prove that for every prime q, h_q = x^5 - 3qx + q has Galois group S_5 over \mathbb{Q}.

Solution

Proof. h_q is Eisenstein at q, so it is irreducible. h_q' = 5x^4 - 3q vanishes only at \pm c with \displaystyle c^4 = \frac{3q}{5}, so h_q is strictly increasing, then decreasing, then increasing, with turning points at -c and c. Since \displaystyle c^4 \ge \frac{6}{5} > 1, c > 1, and

h_q(\pm c) = \pm c(c^4 - 3q) + q = q \mp \tfrac{12q}{5}c,

so \displaystyle h_q(c) < q - \frac{12q}{5} < 0 < h_q(-c). As in Example 3, h_q has exactly three real roots, and Theorem 6 gives S_5. \square

Distinct primes give distinct polynomials, so infinitely many rational quintics are not solvable by radicals.

Boundaries

  • The converse. In characteristic 0 a polynomial with solvable Galois group is solvable by radicals. This is stated without proof; it needs a converse of Theorem 2, proved with Lagrange resolvents. With it, every polynomial of degree at most 4 is solvable by radicals (compare Exercise 11), which is why the cubic and quartic formulas exist.
  • A test, not a characterization. A quintic with one real root can also have group S_5: x^5 - x - 1 does (stated without proof).
  • Prime degree. Theorem 5 fails for 4 and 6 points (Exercises 8 and 9).
  • One formula against all. Corollary 7(b) rules out one formula for every quintic, not formulas for families such as x^5 - a (Example 4).
  • The general polynomial. Over the field \mathbb{Q}(a_0, \dots, a_4) of rational functions in five variables, x^5 + a_4x^4 + \dots + a_0 has Galois group S_5; stated without proof.

Where this goes next

This closes Galois theory in this course. The tower argument with every n_i = 2 is the algebra of Compass-and-straightedge constructions, and why an angle cannot be trisected: a constructible number has degree a power of 2. Over Finite fields every extension has a cyclic group, generated by the Frobenius map. Beyond the course lie the converse of Theorem 4, resolvents that compute the group of a given quintic, and the question of which finite groups occur as Galois groups over \mathbb{Q}.

Exercises

Roots of unity and pure equations

  1. Show that \displaystyle \mathbb{Q}(\omega_8) = \mathbb{Q}(\sqrt{2}, i) and find \operatorname{Gal}(\mathbb{Q}(\omega_8)/\mathbb{Q}) as a subgroup of \mathbb{Z}_8^\times. Is it cyclic?
  2. Given that x^6 + x^3 + 1 is irreducible over \mathbb{Q}, show that \omega_9 is a root of it, and find \operatorname{Gal}(\mathbb{Q}(\omega_9)/\mathbb{Q}) with a generator.
  3. Let F = \mathbb{Q}(\omega_3). Find the Galois group of x^3 - 10 over F, with a generator.
  4. Find the order of the Galois group of x^6 - 5 over \mathbb{Q} and a subnormal series of it with abelian factors.

Quintics and the criterion

  1. Show that 2x^5 - 10x + 5 has Galois group S_5 over \mathbb{Q}.
  2. Show that x^5 - 10x^3 + 5 is not solvable by radicals over \mathbb{Q}.
  3. Let g = x(x^2 + 2)(x^2 - 4)(x^2 - 16) + 2, which expands to x^7 - 18x^5 + 24x^3 + 128x + 2. Show that its Galois group over \mathbb{Q} is S_7.
  4. Show that (1\ 2\ 3\ 4) and (1\ 3) generate a subgroup of S_4 of order 8. Why does this not contradict Theorem 5?
  5. Jin and Amara are building a puzzle box whose lid has six sliding tiles set in a ring at positions 1 to 6. One lever turns the ring by \rho = (1\ 2\ 3\ 4\ 5\ 6); the other swaps the tiles at positions 1 and 4. (a) Show that tiles that start opposite each other stay opposite, and that exactly 24 arrangements can be reached. (b) Amara rewires the second lever to swap positions 1 and 2. How many arrangements can be reached now?

Proofs

  1. Prove that an irreducible f \in \mathbb{Q}[x] of degree 5 that is solvable by radicals over \mathbb{Q} has exactly one or exactly five real roots.
  2. Prove that every nonconstant f \in F[x] of degree at most 4 has a solvable Galois group over F.
  3. Let p be prime and G \le S_p act transitively on \{1, \dots, p\}. Prove that if G contains a transposition, then G = S_p.
  4. Let p be prime, F a field containing \omega_p, and a \in F. Prove that x^p - a is either irreducible over F or a product of linear factors over F.
  5. Let n be a positive integer. Prove that every field K with \mathbb{Q} \subseteq K \subseteq \mathbb{Q}(\omega_n) is Galois over \mathbb{Q} with abelian Galois group.
  6. Let F \subseteq K be fields and f \in F[x] solvable by radicals over F. Prove that f is solvable by radicals over K.
Answers
  1. \displaystyle \omega_8 = \frac{\sqrt{2}}{2} + \frac{\sqrt{2}}{2}i; conversely i = \omega_8^2 and \displaystyle \sqrt{2} = \omega_8 + \omega_8^{-1}. The degree is 4 (\displaystyle \sqrt{2} is real, i is not), so Theorem 1’s image is all of \mathbb{Z}_8^\times = \{1, 3, 5, 7\}. As 3^2, 5^2 and 7^2 are \equiv 1 \pmod 8, the group is V_4, not cyclic.
  2. \omega_9^3 \ne 1 and x^9 - 1 = (x^3 - 1)(x^6 + x^3 + 1), so \omega_9 is a root of the second factor, and the degree is 6. Theorem 1’s image is all of \mathbb{Z}_9^\times. The powers of 2 modulo 9 are 2, 4, 8, 7, 5, 1, so the group is cyclic of order 6, generated by \sigma_2 : \omega_9 \mapsto \omega_9^2.
  3. Let \displaystyle \beta = \sqrt[3]{10}, real; by Theorem 2 the splitting field is F(\beta). x^3 - 10 is Eisenstein at 5, so [\mathbb{Q}(\beta) : \mathbb{Q}] = 3, and [F : \mathbb{Q}] = 2; these are coprime, so [F(\beta) : F] = 3. The image in \mu_3 has 3 elements: the group is cyclic of order 3, generated by \beta \mapsto \omega_3\beta.
  4. Let \displaystyle \beta = \sqrt[6]{5}, real; the roots are \beta\omega_6^k, and E = \mathbb{Q}(\omega_6, \beta) with \displaystyle \mathbb{Q}(\omega_6) = \mathbb{Q}(\sqrt{3}\,i) of degree 2. x^6 - 5 is Eisenstein at 5 and \displaystyle \sqrt{3}\,i \notin \mathbb{Q}(\beta) \subseteq \mathbb{R}, so [E : \mathbb{Q}] = 12, the order. H = \operatorname{Gal}(E/\mathbb{Q}(\omega_6)) is cyclic of order 6 (Theorem 2) and normal, so G \ge H \ge \{e\} has factors of orders 2 and 6.
  5. Eisenstein at 5: 5 \nmid 2, 5 divides 10 and 5, 25 \nmid 5. f' = 10x^4 - 10 vanishes at \pm 1; f(1) = -3 < 0 and f(-1) = 13 > 0, so f rises, falls, rises, with three real roots. Theorem 6 gives S_5.
  6. Eisenstein at 5. f' = 5x^2(x^2 - 6) is negative on \displaystyle (-\sqrt{6}, \sqrt{6}) except at 0 and positive outside. \displaystyle f(\sqrt{6}) = 5 - 24\sqrt{6} < 0 and \displaystyle f(-\sqrt{6}) = 5 + 24\sqrt{6} > 0, so there are exactly three real roots, and Corollary 7(a) applies.
  7. Eisenstein at 2. The values g(-5) = -25513, g(-3) = 1157, g(-1) = -133, g(1) = 137, g(3) = -1153, g(5) = 25517 alternate in sign, giving at least five real roots. g' = P(x^2) with P(y) = 7y^3 - 90y^2 + 72y + 128; P(0) = 128, P(2) = -32, P(12) = 128 give two positive roots of P, and the product of its roots is \displaystyle -\frac{128}{7} < 0, so the third is negative. So g' has exactly four real roots, and by Rolle’s theorem g, whose roots are distinct, has at most five. Two roots are non-real, and Theorem 6 with p = 7 gives S_7.
  8. Number the corners of a square 1, 2, 3, 4 in order: (1\ 2\ 3\ 4) is a rotation and (1\ 3) a reflection, so both lie in the copy of D_4, of order 8, acting on the corners. The subgroup contains \langle (1\ 2\ 3\ 4) \rangle, of order 4, and (1\ 3) outside it, so its order is 8 (Lagrange's theorem). No contradiction: 4 is not prime, and in the proof of Theorem 5 the power (1\ 3)(2\ 4) carries 1 to 3 without being a 4-cycle.
  9. (a) Both levers permute the opposite pairs \{1, 4\}, \{2, 5\}, \{3, 6\}, so opposite tiles stay opposite. The group G of the levers maps onto \langle (1\ 2\ 3) \rangle, its action on the three pairs (the swap fixes each pair). The kernel lies in \langle (1\ 4), (2\ 5), (3\ 6) \rangle, of order 8, and contains all of it, since \rho(1\ 4)\rho^{-1} = (2\ 5) and \rho^2(1\ 4)\rho^{-2} = (3\ 6). So |G| = 8 \cdot 3 = 24 (first isomorphism theorem), one arrangement per element. (b) Conjugating (1\ 2) by powers of \rho gives (2\ 3), (3\ 4), (4\ 5), (5\ 6), which generate S_6 (proof of Theorem 5): all 720 arrangements.
  10. Proof. The five roots are distinct (Field automorphisms and the Galois group, Theorem 6), and the non-real ones come in conjugate pairs (Complex roots come in pairs), so f has 5, 3 or 1 real roots. With 3, Corollary 7(a) says f is not solvable by radicals. So f has 1 or 5. \square
  11. Proof. By Field automorphisms and the Galois group, Theorem 2(b), the group is isomorphic to a subgroup of S_r, where r \le 4 is the number of distinct roots. S_1 and S_2 are abelian, and S_3 \ge A_3 \ge \{e\} and S_4 \ge A_4 \ge V_4 \ge \{e\} have abelian factors (Solvable groups, Example 2), so S_r is solvable; so is every subgroup (Theorem 4(a) there), and so is any group isomorphic to one (Lemma 2(a) there). \square
  12. Proof. The orbit of 1 has p points, so p divides |G| by the orbit-stabilizer theorem (Group actions, orbits and stabilizers). A Sylow p-subgroup is then nontrivial (The Sylow theorems), and a power of one of its elements other than e has order p. A permutation of p points of order p has cycle lengths with least common multiple p, so it is a p-cycle. Theorem 5 gives G = S_p. \square
  13. Proof. If a = 0, x^p is a product of linear factors. Otherwise let \beta^p = a. By Theorem 2, F(\beta) is the splitting field and its group embeds in \mu_p, so it has order 1 or p, and [F(\beta) : F] equals that order (Field automorphisms and the Galois group, Corollary 7). If it is 1, \beta \in F and every root \beta\omega_p^k lies in F. If it is p, m_\beta has degree p and divides x^p - a; both are monic, so they are equal and x^p - a is irreducible. \square
  14. Proof. \mathbb{Q}(\omega_n)/\mathbb{Q} is Galois with abelian group G (Theorem 1). K corresponds to H = \operatorname{Gal}(\mathbb{Q}(\omega_n)/K), normal because every subgroup of an abelian group is. By The fundamental theorem of Galois theory, Theorem 5(d), K/\mathbb{Q} is Galois with group G/H, a quotient of an abelian group, so abelian. \square
  15. Proof. Let the splitting field of f over F lie in L = F(\alpha_1, \dots, \alpha_m) with \alpha_i^{n_i} \in F(\alpha_1, \dots, \alpha_{i-1}). Then L' = K(\alpha_1, \dots, \alpha_m) is radical over K, since F(\alpha_1, \dots, \alpha_{i-1}) \subseteq K(\alpha_1, \dots, \alpha_{i-1}). The roots of f lie in L \subseteq L', so the splitting field of f over K, which they generate over K, lies in L'. \square
math.wiki · updated 2026-11-18 · CC-BY-SA-4.0