Solvable groups
Which groups can be taken apart, one normal subgroup at a time, into pieces that are all abelian, and why does the symmetric group on five letters refuse?
Where you are 4 in Galois theory
Why this exists
The quadratic formula solves every quadratic with a square root, and formulas built from square and cube roots solve every cubic and quartic. A formula in radicals adjoins one root \displaystyle \sqrt[n]{a} after another, climbing a tower of fields. With the right roots of unity present, each step contributes an abelian layer of symmetry, and the fundamental theorem of Galois theory turns the tower into a chain of subgroups of the Galois group, each normal in the one before with an abelian quotient. So solving by radicals becomes a question about one finite group: can it be broken into abelian pieces? This lesson proves that S_n can for n \le 4 and cannot for n \ge 5, the group-theory half of why the quintic has no formula.
Definition
In plain words: a group is solvable when a finite chain of subgroups, each normal in the one before, leads from the group down to \{e\} with every step’s quotient abelian.
- For a, b in a group G, the commutator is [a, b] = aba^{-1}b^{-1}. It equals e exactly when ab = ba. The commutator subgroup [G, G] is the subgroup generated by all commutators. Since [a, b]^{-1} = bab^{-1}a^{-1} = [b, a] and e = [e, e], its elements are exactly the finite products of commutators.
- The derived series is G^{(0)} = G, G^{(i+1)} = [G^{(i)}, G^{(i)}].
- A subnormal series of G is a chain of subgroups G = G_0 \ge G_1 \ge \dots \ge G_n = \{e\} with G_{i+1} \trianglelefteq G_i for each i. Its factors are the quotients G_i/G_{i+1}, and n is its length.
- G is solvable if it has a subnormal series whose factors are all abelian. If G^{(n)} = \{e\} for some n, the least such n is the derived length of G.
- A group G \ne \{e\} is simple if its only normal subgroups are \{e\} and G.
An abelian group is solvable through G \ge \{e\}. The group S_3 is solvable without being abelian: in S_3 \ge A_3 \ge \{e\} the factors have orders 2 and 3.
Intuition
A factor G_i/G_{i+1} is what one step removes. The commutator subgroup is the smallest normal subgroup with an abelian quotient (Theorem 1), so G^{(1)} is as low as one abelian step can reach, G^{(2)} as low as two steps reach, and so on: the derived series is the fastest descent, and G is solvable exactly when it reaches the bottom (Theorem 3). In S_n with n \ge 5, every 3-cycle is a commutator of two 3-cycles, so no abelian step ever removes a 3-cycle, and the descent stalls at A_n.
Results
Throughout, G is a group.
Theorem 1 (the commutator subgroup). (a) [G, G] \trianglelefteq G. (b) For N \trianglelefteq G, G/N is abelian if and only if [G, G] \le N. In particular G/[G, G] is abelian, and [G, G] = \{e\} if and only if G is abelian.
Lemma 2 (images and subgroups). Let \psi : G \to L be a homomorphism and H \le G. For every i \ge 0: (a) \psi(G^{(i)}) = \psi(G)^{(i)}; (b) H^{(i)} \le G^{(i)}; (c) G^{(i)} \trianglelefteq G.
Theorem 3 (the derived series test). If G = G_0 \ge G_1 \ge \dots \ge G_n = \{e\} is a subnormal series with abelian factors, then G^{(i)} \le G_i for every i. Hence G is solvable if and only if G^{(n)} = \{e\} for some n, and then the derived length is the least length of a subnormal series with abelian factors.
Theorem 4 (subgroups, quotients, extensions). Let N \trianglelefteq G. (a) If G is solvable, so is every subgroup of G. (b) If G is solvable, so is every homomorphic image of G, in particular G/N. (c) If N and G/N are solvable, so is G.
Corollary 5. (a) Every abelian group is solvable. (b) If p is prime, every group of order p^k is solvable.
Theorem 6 (prime factors). A finite group is solvable if and only if it has a subnormal series whose factors are all cyclic of prime order.
Lemma 7 (three-cycles). (a) For n \ge 3, every element of A_n is a product of 3-cycles. (b) If i, j, k, l, m \in \{1, \dots, n\} are distinct, then (i\ j\ k) = [(i\ j\ l), (i\ k\ m)].
Theorem 8 (three-cycles survive abelian steps). Let n \ge 5, let G \le S_n contain every 3-cycle, and let N \trianglelefteq G with G/N abelian. Then N contains every 3-cycle.
Corollary 9 (symmetric groups). Let n \ge 5. Then [S_n, S_n] = A_n and [A_n, A_n] = A_n, so S_n^{(i)} = A_n for every i \ge 1. Neither S_n nor A_n is solvable, and neither is any group with a subgroup isomorphic to A_5. For n \le 4, S_n is solvable.
Why it is true
Proof of Theorem 1. (a) Conjugation c_g(x) = gxg^{-1} satisfies c_g(xy) = gxg^{-1}gyg^{-1} = c_g(x)c_g(y), so it is a homomorphism, and c_g([a, b]) = [c_g(a), c_g(b)]. Thus c_g sends products of commutators to products of commutators: g[G, G]g^{-1} \subseteq [G, G] for every g, and [G, G] \trianglelefteq G (Normal subgroups, Theorem 2). (b) By Quotient groups, Theorem 6(c), G/N is abelian if and only if N contains every commutator, and a subgroup contains every commutator exactly when it contains the subgroup they generate. Taking N = [G, G] gives the next claim, and [G, G] = \{e\} says every commutator is e, that is, all elements commute. \square
Proof of Lemma 2. (a) \psi([a, b]) = [\psi(a), \psi(b)], and every commutator of \psi(G) has this form, so \psi maps the commutators of G onto those of \psi(G), and products of them onto products of them: \psi([G, G]) = [\psi(G), \psi(G)]. Applied to \psi restricted to G^{(i)}, with \psi(G^{(i)}) = \psi(G)^{(i)}, this gives \psi(G^{(i+1)}) = \psi(G)^{(i+1)}, the step of a Proof by induction. (b) If K \le M, every commutator of K is one of M, so [K, K] \le [M, M]; induction on i gives (b). (c) Apply (a) to c_g : G \to G, which is onto: gG^{(i)}g^{-1} = c_g(G)^{(i)} = G^{(i)}. \square
Proof of Theorem 3. Induction on i, with G^{(0)} = G_0. If G^{(i)} \le G_i: as G_i/G_{i+1} is abelian, Theorem 1(b) in the group G_i gives [G_i, G_i] \le G_{i+1}, and Lemma 2(b) gives G^{(i+1)} = [G^{(i)}, G^{(i)}] \le [G_i, G_i] \le G_{i+1}. So G^{(n)} \le G_n = \{e\}. Conversely, if G^{(n)} = \{e\}, the derived series G^{(0)} \ge \dots \ge G^{(n)} is a subnormal series (Theorem 1(a) in each G^{(i)}) with abelian factors (Theorem 1(b)) and length n; for the least such n, the first part rules out a shorter series. \square
Proof of Theorem 4. (a) and (b): let G^{(n)} = \{e\} (Theorem 3). Then H^{(n)} \le G^{(n)} = \{e\} by Lemma 2(b), and \psi(G)^{(n)} = \psi(G^{(n)}) = \{e\} by Lemma 2(a); the natural map \pi : G \to G/N gives G/N. (c) Let (G/N)^{(m)} = \{N\} and N^{(k)} = \{e\}. Lemma 2(a) gives \pi(G^{(m)}) = (G/N)^{(m)} = \{N\}, so G^{(m)} \le N. From the definition, G^{(m+k)} = (G^{(m)})^{(k)}, and Lemma 2(b) bounds it by N^{(k)} = \{e\}. \square
Proof of Corollary 5. (a) [G, G] = \{e\}. (b) Strong induction and well-ordering on k; for k = 0, G = \{e\}. For k \ge 1, Z(G) \ne \{e\} (Group actions, orbits and stabilizers, Corollary 7), Z(G) is abelian and normal, and G/Z(G) has order p^j with j < k, so it is solvable. Theorem 4(c) finishes. \square
Proof of Theorem 6. Cyclic groups are abelian, so one direction is the definition. For the other, refine each step of a series with abelian factors, using this claim: if M \trianglelefteq K and K/M is finite and abelian, there is a chain K = K_0 \ge K_1 \ge \dots \ge K_r = M with each K_{j+1} \trianglelefteq K_j and each factor cyclic of prime order. Strong induction on |K/M|; for |K/M| = 1 take r = 0. Otherwise choose xM \ne M, of order m > 1, and a prime p \mid m; then yM = (xM)^{m/p} has order p. By the correspondence theorem (The isomorphism theorems, Theorem 5, in the group K), \langle yM \rangle = L/M with M \le L \le K, and L \trianglelefteq K because every subgroup of the abelian group K/M is normal. By the third isomorphism theorem (Theorem 6 there), K/L \cong (K/M)/(L/M), abelian of order |K/M|/p, so induction gives a chain from K down to L. Append L \ge M: M \trianglelefteq L, and L/M has prime order, so it is cyclic (Lagrange's theorem). \square
Proof of Lemma 7. (a) By Permutation groups and cycle notation, Theorems 5 and 7, an even permutation is a product of an even number of transpositions. Group them in pairs. A pair is (i\ j)(i\ j) = e, or shares one point, (i\ j)(i\ k) = (i\ k\ j), or is disjoint, (i\ j)(k\ l) = (i\ k\ j)(i\ k\ l), each checked by following points.
(b) First let x = (1\ 2\ 4) and y = (1\ 3\ 5). In xyx^{-1}y^{-1} the factor y^{-1} = (1\ 5\ 3) acts first, then x^{-1} = (1\ 4\ 2), then y, then x. Following each point:
So [x, y] = (1\ 2\ 3). In general, let \pi \in S_n send 1, 2, 3, 4, 5 to i, j, k, l, m. Conjugation relabels cycles, \pi(a_1\ \cdots\ a_r)\pi^{-1} = (\pi(a_1)\ \cdots\ \pi(a_r)) (Normal subgroups, Example 4), and c_\pi preserves commutators (proof of Theorem 1), so (i\ j\ k) = c_\pi([x, y]) = [(i\ j\ l), (i\ k\ m)]. \square
Proof of Theorem 8. Let (i\ j\ k) be a 3-cycle. As n \ge 5, there are distinct l, m outside \{i, j, k\}. The 3-cycles x = (i\ j\ l) and y = (i\ k\ m) lie in G, and G/N is abelian, so [x, y] \in N (Theorem 1(b)). By Lemma 7(b), [x, y] = (i\ j\ k). \square
Proof of Corollary 9. The sign is a homomorphism to the abelian group \{\pm 1\} (Permutation groups and cycle notation, Theorem 7(b)), so \operatorname{sgn}([a, b]) = 1 and [S_n, S_n] \le A_n. Theorem 8 with G = S_n and N = [S_n, S_n], whose quotient is abelian by Theorem 1(b), puts every 3-cycle in [S_n, S_n], so Lemma 7(a) gives A_n \le [S_n, S_n]. The 3-cycles are even, so A_n contains them all, and the same argument with G = A_n gives [A_n, A_n] = A_n. By induction S_n^{(i)} = A_n \ne \{e\} for i \ge 1 and A_n^{(i)} = A_n, so neither is solvable (Theorem 3). An isomorphic copy of A_5 is not solvable either, by Lemma 2(a) applied to an isomorphism, so no group containing one is (Theorem 4(a)). For n \le 4: S_1 and S_2 are abelian, S_3 was done above, and S_4 is Example 2. \square
Worked examples
Commutators in a dihedral group
Let n \ge 3. Find [D_n, D_n] and the derived length of D_n.
SolutionFrom srs = r^{-1}, inverting both sides gives sr^{-1}s = r, so
Let N = \langle r^2 \rangle. Since rr^2r^{-1} = r^2 and sr^2s^{-1} = r^{-2} \in N, checking on generators gives N \trianglelefteq D_n (Normal subgroups, Theorem 3). The quotient is generated by rN and sN, and these commute: rN \cdot sN = rsN and sN \cdot rN = srN = r^{-1}sN, which are equal because (r^{-1}s)^{-1}(rs) = s r^2 s = r^{-2} \in N. So D_n/N is abelian, Theorem 1(b) gives [D_n, D_n] \le N, and r^2 \in [D_n, D_n] gives equality: [D_n, D_n] = \langle r^2 \rangle.
This group is cyclic, so D_n^{(2)} = \{e\}; and D_n^{(1)} \ne \{e\} because rs = sr would force r = r^{-1}, false for n \ge 3. The derived length is 2. For odd n, \langle r^2 \rangle = \langle r \rangle has order n and the quotient has order 2; for even n it has order n/2 and the quotient has order 4. In D_6, [D_6, D_6] = \{e, r^2, r^4\}.
The derived series of S_4
Find the derived series of S_4, with V = \{e, (1\ 2)(3\ 4), (1\ 3)(2\ 4), (1\ 4)(2\ 3)\}.
SolutionCommutators are even (proof of Corollary 9), so [S_4, S_4] \le A_4. Transpositions are their own inverses, so
[S_4, S_4] is normal (Theorem 1(a)) and conjugation relabels, so it holds all eight 3-cycles and e: at least 9 elements of A_4, with order dividing 12 (Lagrange's theorem). So [S_4, S_4] = A_4.
Next, V \trianglelefteq A_4 (Normal subgroups, Exercise 8) and A_4/V has prime order 3, so it is abelian and [A_4, A_4] \le V. Following points as in Lemma 7, [(1\ 2\ 3), (1\ 2\ 4)] = (1\ 2)(3\ 4). Conjugating by (1\ 2\ 3) and by (1\ 3\ 2) relabels this to (2\ 3)(1\ 4) and (3\ 1)(2\ 4), and [A_4, A_4] \trianglelefteq A_4, so [A_4, A_4] = V. Finally V is abelian. The derived series is
with factors of orders 2, 3 and 4, and S_4 has derived length 3.
A solvable group of order 20 inside S_5
Let \sigma = (1\ 2\ 3\ 4\ 5) and \tau = (2\ 3\ 5\ 4) in S_5, and G = \langle \sigma, \tau \rangle. Show that |G| = 20, find the derived series, and give a subnormal series whose factors have prime order.
SolutionRelabeling, \tau\sigma\tau^{-1} = (1\ 3\ 5\ 2\ 4), which is \sigma^2. So P = \langle \sigma \rangle, of order 5, is normalized by both generators, and P \trianglelefteq G (Normal subgroups, Theorem 3). Let T = \langle \tau \rangle, of order 4. Then TP is a subgroup (Theorem 6 there) containing \sigma and \tau, so TP = G. Every power of \tau fixes 1 and no nontrivial power of \sigma does, so T \cap P = \{e\}, and the second isomorphism theorem (The isomorphism theorems, Theorem 4) gives G/P \cong T/(T \cap P) \cong T, cyclic of order 4. Hence |G| = 5 \cdot 4 = 20.
Since G/P is abelian, [G, G] \le P (Theorem 1(b)); and [\tau, \sigma] = \tau\sigma\tau^{-1}\sigma^{-1} = \sigma^2\sigma^{-1} = \sigma, so [G, G] = P. As P is abelian, the derived series is G \ge P \ge \{e\}, with factors cyclic of orders 4 and 5.
For prime factors, refine the factor of order 4 as in Theorem 6: its subgroup of order 2 is generated by \tau^2P, with \tau^2 = (2\ 5)(3\ 4), and corresponds to H = \langle \sigma, \tau^2 \rangle of order 10, normal in G because G/P is abelian. The series G \ge H \ge P \ge \{e\} has factors of orders 2, 2, 5.
Groups of order pq and p^2q
Let p and q be distinct primes. Prove that every group of order pq and every group of order p^2q is solvable.
SolutionProof. Let |G| = pq, and name the primes so that p < q. By the third Sylow theorem (The Sylow theorems, Theorem 4), n_q divides p and n_q \equiv 1 \pmod{q}; as 1 < p < q, n_q = 1. So the Sylow q-subgroup Q is normal (Corollary 5 there). Q and G/Q have prime orders q and p, so G \ge Q \ge \{e\} has cyclic factors.
Now let |G| = p^2q. By The Sylow theorems, Exercise 17, G is not simple, so some N \trianglelefteq G has \{e\} \ne N \ne G. Then |N| and |G/N| are divisors of p^2q other than 1 and p^2q, namely p, p^2, q or pq. Groups of the first three orders are solvable by Corollary 5(b), and of order pq by the first part. Theorem 4(c) gives G solvable. \square
Upper unitriangular matrices
Let F be a field and U the group of matrices
Find the derived series of U.
SolutionMultiplying out,
so U is closed under products, u(0, 0, 0) = I_3, and u(a, b, c)^{-1} = u(-a, -b, ab - c). The formula shows that \psi(u(a, b, c)) = (a, b) is a homomorphism onto the additive group F \times F, with kernel Z = \{u(0, 0, c) : c \in F\}. By the first isomorphism theorem (The isomorphism theorems, Theorem 2), U/Z \cong F \times F is abelian, so [U, U] \le Z.
For the other inclusion, u(a, 0, 0)u(0, b, 0) = u(a, b, ab) and u(0, b, 0)u(a, 0, 0) = u(a, b, 0), so
Taking b = 1, every element of Z is a commutator, so [U, U] = Z. As Z \cong F under addition is abelian, U^{(2)} = \{I_3\}, and U is not abelian since u(0, 0, 1) \ne I_3 is a commutator. The derived length is 2.
Boundaries
- Small n. Lemma 7(b) needs five distinct points. In A_4 no 3-cycle is a commutator of elements of A_4, since [A_4, A_4] = V (Example 2).
- Normal in the previous term only. A term need not be normal in G: in Exercise 4, \langle (1\ 2)(3\ 4) \rangle is not normal in S_4. The terms of the derived series are (Lemma 2(c)).
- Infinite groups. \mathbb{Z} is solvable, but Theorem 6 fails for it: every subgroup k\mathbb{Z} \ne \{0\} is infinite, so no factor next to \{0\} has prime order.
- Beyond this lesson. A_5 is in fact simple, and every group of order less than 60 is solvable; neither is proved here.
Where this goes next
Why the quintic has no formula uses the The fundamental theorem of Galois theory to show that, in characteristic 0, a polynomial solvable by radicals has a solvable Galois group, then exhibits a rational quintic whose Galois group is S_5, which Corollary 9 rules out. Every solvable subgroup of S_5 that moves 1 to every point lies in a conjugate of the group of order 20 in Example 3 (a theorem of Galois, not proved here).
Exercises
Commutators and derived series
- In S_5, compute [(1\ 4)(2\ 5), (1\ 2\ 3)].
- Find the derived series and derived length of S_3 \times D_4.
- Let B be the group of invertible upper triangular 2 \times 2 matrices over \mathbb{Z}_3, writing k for [k]. Find |B| and the derived series of B.
- Refine the derived series of S_4 (Example 2) to a subnormal series whose factors have prime order, and list the orders of the factors.
Three-cycles and symmetric groups
- Use Lemma 7(b) to write (1\ 4\ 5) as a commutator of two 3-cycles in A_5, and check it by following points.
- Following the proof of Lemma 7(a), write (1\ 2)(4\ 5) and (1\ 3)(1\ 5) as products of 3-cycles.
- Yuna and Andre ring five handbells set at five numbered places on a table. Between rounds they may apply the shift \rho = (1\ 5\ 4\ 3\ 2), which moves the bell at place 1 to place 5, and so on, or the swap (2\ 3). Show that every arrangement of the bells can be reached by repeating these two moves, and decide whether the group they generate is solvable.
- Decide which are solvable: (a) D_9 \times A_4; (b) A_6; (c) S_4 \times S_4; (d) a group of order 75; (e) the subgroup of S_7 of permutations fixing 6 and 7.
Proofs
- Prove that G \times H is solvable if and only if G and H are.
- Prove that a simple solvable group is cyclic of prime order.
- Let N \trianglelefteq G with N \cap [G, G] = \{e\}. Prove that N \le Z(G).
- Prove that a finite solvable group G \ne \{e\} has a normal subgroup of prime index.
- Let n \ge 5. Prove that the only normal subgroups N of S_n with S_n/N abelian are A_n and S_n.
- Prove that every group of order 12 is solvable.
Answers
- Write x = (1\ 4)(2\ 5), y = (1\ 2\ 3). Relabeling, xyx^{-1} = (4\ 5\ 3), so [x, y] = (4\ 5\ 3)(1\ 3\ 2). Following points, right factor first: 1 \mapsto 3 \mapsto 4, 4 \mapsto 5, 5 \mapsto 3, 3 \mapsto 2, 2 \mapsto 1. So [x, y] = (1\ 4\ 5\ 3\ 2), even, as every commutator is.
- In a product, [(a, b), (c, d)] = ([a, c], [b, d]), so the commutators of G \times H are the pairs of commutators, and since [G \times H, G \times H] contains each (c, e) and (e, c'), it is [G, G] \times [H, H]. Here [S_3, S_3] = A_3: commutators are even and [(1\ 3), (2\ 3)] = ((1\ 3)(2\ 3))^2 = (1\ 3\ 2)^2 = (1\ 2\ 3). By Example 1, [D_4, D_4] = \{e, r^2\}. So the first derived group is A_3 \times \{e, r^2\}, of order 6 and abelian, the next is trivial, and the derived length is 2.
- An invertible upper triangular matrix has diagonal entries a, d \in \{1, 2\} and any corner b \in \mathbb{Z}_3, so |B| = 2 \cdot 2 \cdot 3 = 12. Sending a matrix to its diagonal pair (a, d) is a homomorphism onto the abelian group \mathbb{Z}_3^\times \times \mathbb{Z}_3^\times with kernel U = \{U_0, U_1, U_2\}, U_t = \begin{pmatrix} 1 & t \\ 0 & 1 \end{pmatrix}, so [B, B] \le U. With g = \begin{pmatrix} 2 & 0 \\ 0 & 1 \end{pmatrix}, gU_1g^{-1} = U_2 (Normal subgroups, Example 2), so [g, U_1] = U_2U_{-1} = U_1 \ne I_2. Hence [B, B] = U \cong \mathbb{Z}_3, abelian, and the series is B \ge U \ge \{I_2\}, derived length 2.
- V is abelian, so its subgroup W = \langle (1\ 2)(3\ 4) \rangle is normal in V. The series S_4 \ge A_4 \ge V \ge W \ge \{e\} has factors of orders 2, 3, 2, 2.
- With i, j, k = 1, 4, 5 and l, m = 2, 3: (1\ 4\ 5) = [(1\ 4\ 2), (1\ 5\ 3)]. Here y^{-1} = (1\ 3\ 5) acts first, then x^{-1} = (1\ 2\ 4), y, x: 1 \mapsto 3 \mapsto 3 \mapsto 1 \mapsto 4; 4 \mapsto 4 \mapsto 1 \mapsto 5 \mapsto 5; 5 \mapsto 1 \mapsto 2 \mapsto 2 \mapsto 1; 2 \mapsto 2 \mapsto 4 \mapsto 4 \mapsto 2; 3 \mapsto 5 \mapsto 5 \mapsto 3 \mapsto 3.
- Disjoint pair with i, j, k, l = 1, 2, 4, 5: (1\ 2)(4\ 5) = (1\ 4\ 2)(1\ 4\ 5). Shared point with i, j, k = 1, 3, 5: (1\ 3)(1\ 5) = (1\ 5\ 3).
- Let G = \langle \rho, (2\ 3) \rangle; in a finite group each inverse is a positive power, so products of moves already form G. Conjugating relabels, \rho^k(2\ 3)\rho^{-k} = (\rho^k(2)\ \rho^k(3)), which for k = 1, 2, 3, 4 gives (1\ 2), (5\ 1), (4\ 5), (3\ 4). From (1\ 2), (2\ 3), (3\ 4), (4\ 5) every transposition follows, by induction on j - i: (i\ j) = (j{-}1\ \ j)(i\ \ j{-}1)(j{-}1\ \ j). Every permutation is a product of transpositions (Permutation groups and cycle notation, Theorem 5), so G = S_5 and every arrangement is reachable. By Corollary 9, G is not solvable.
- (a) Solvable: D_9 by Example 1, A_4 by Example 2 and Theorem 4(a), and the product by Exercise 9. (b) Not solvable, by Corollary 9. (c) Solvable, by Example 2 and Exercise 9. (d) Solvable: 75 = 5^2 \cdot 3, so Example 4 applies. (e) Not solvable: restriction to \{1, \dots, 5\} is an isomorphism onto S_5 (Corollary 9 and Lemma 2(a)).
- Proof. The projection (g, h) \mapsto g is a homomorphism onto G, so if G \times H is solvable, so are G and likewise H (Theorem 4(b)). Conversely, N = \{e\} \times H is the kernel of that projection, so N \trianglelefteq G \times H and (G \times H)/N \cong G (The isomorphism theorems, Theorem 2); and (e, h) \mapsto h is an isomorphism N \to H. Both are solvable by Lemma 2(a), so G \times H is solvable by Theorem 4(c). \square
- Proof. Let G be simple and solvable; then G \ne \{e\}. As [G, G] \trianglelefteq G, it is \{e\} or G. If [G, G] = G, then G^{(i)} = G \ne \{e\} for every i, contradicting Theorem 3. So G is abelian and every subgroup is normal. For g \ne e, \langle g \rangle = G. If g had infinite order, \langle g^2 \rangle would be normal and neither \{e\} nor G (as g = g^{2t} would give g^{2t - 1} = e). So \operatorname{ord}(g) = m is finite; for a prime p \mid m, the normal subgroup \langle g^p \rangle has order m/p < m, so m/p = 1 and |G| = p. \square
- Proof. Let x \in N and g \in G. Then [x, g] = x(gx^{-1}g^{-1}) lies in N, as gx^{-1}g^{-1} \in N by normality, and in [G, G]. So [x, g] = e, that is xg = gx, for every g, and x \in Z(G). \square
- Proof. If [G, G] = G, then G^{(i)} = G \ne \{e\} for all i, against Theorem 3; so A = G/[G, G] is a nontrivial finite abelian group. The claim in the proof of Theorem 6, with K = A and M = \{e\}, gives A \ge A_1 with |A/A_1| = p prime. By the correspondence theorem (The isomorphism theorems, Theorem 5), A_1 = N/[G, G] for some N with [G, G] \le N \le G; A_1 \trianglelefteq A as A is abelian, so N \trianglelefteq G, and [G : N] = [A : A_1] = p. \square
- Proof. If S_n/N is abelian, then A_n = [S_n, S_n] \le N (Theorem 1(b), Corollary 9). As [S_n : A_n] = 2, Lagrange’s theorem leaves N = A_n or N = S_n. Both qualify: S_n/A_n has order 2, and S_n/S_n is trivial. \square
- Proof. Let |G| = 12 = 2^2 \cdot 3. By The Sylow theorems, Exercise 8(a), n_3 = 1 or n_2 = 1, so a Sylow subgroup P of order 3 or 4 is normal (Corollary 5 there). Then |G/P| is 4 or 3. Groups of orders 3 and 4 have prime power order, so P and G/P are solvable (Corollary 5(b)), and Theorem 4(c) gives G solvable. \square