Notes on Abstract Algebra

John Peloquin

Fall 2007

Abstract
Notes on Abstract Algebra by Dummit and Foote.

Exercises
Chapter 0
Section 0.3

E XERCISE 12–14. Let n ∈ Z, n ≥ 1. Then

(Z/nZ)× = { a ∈ Z/nZ | (a, n) = 1 }

We first note that the set on the right is well defined. For suppose a, b ∈ Z and a ≡ b
mod n, so a = b + nk for some k ∈ Z. If (b, n) = d > 1, write b = d k 1 and n = d k 2 ,
where k 1 , k 2 ∈ Z. Then

a = b + nk = d k 1 + (d k 2 )k = d (k 1 + k 2 k)

so d |a, and (a, n) > 1. Hence if (a, n) = 1, then (b, n) = 1. Since b was arbitrary, this
shows that the set is well defined. Now for the proof.

Proof. The case n = 1 is trivial, so we assume n > 1.

First, suppose 1 ≤ a < n and (a, n) = 1. Then we know (Euclidean algorithm)
that there exist x, y ∈ Z such that ax + n y = 1. In other words, ax = 1 + n(−y), so
ax = a · x = 1. Thus a ∈ (Z/nZ)× .
Conversely, suppose 1 ≤ a < n and (a, n) = d > 1. Write a = d k 1 and n = d k 2 ,
where k 1 , k 2 ∈ Z and 1 ≤ k 2 < n. Then ak 2 = d k 1 k 2 = nk 1 , so a · k 2 = 0. Now k 2 6= 0.
But if there exists b ∈ Z such that a · b = 1, then

k 2 = 1 · k 2 = (a · b) · k 2 = b · (a · k 2 ) = ·k 2 = b · 0 = 0

—a contradiction. Thus a 6∈ (Z/nZ)× .

It follows from the above proof that Z/nZ is partitioned into two subsets, one
consisting of all elements with multiplicative inverses and the other consisting of all
elements which are zero divisors. (Recall that a ∈ Z/nZ is called a zero divisor iff
a 6= 0 and there exists b ∈ Z/nZ, b 6= 0 such that ab = 0.)

1
Chapter 1
Section 1

E XERCISE 5. Let n > 1. Then Z/nZ is not a group under multiplication.

Proof. Note that 0, 1 ∈ Z/nZ and 0 6= 1. If Z/nZ were a group under multiplication,
then there would exist a ∈ Z/nZ such that

1 = 0·a = 0·a = 0

—a contradiction.

E XERCISE 25. Let G be a group and suppose x 2 = 1 for all x ∈ G. Then G is abelian.

Proof. Let a, b ∈ G. Then

ab = (ba)2 ab = bab(a 2 )b = ba(b 2 ) = ba

E XERCISE 26. Let G be a group and x ∈ G. Set H = { x n | n ∈ Z }. Then H is a subgroup

of G.

Proof. Note that x 0 = e ∈ H . If x m , x n ∈ H , then x m · x n = x m+n ∈ H , and x −m ∈ H .

Thus H is nonempty and closed under products and inverses, and so is a subgroup.

We call H the cyclic subgroup generated by x and write H = 〈x〉.

E XERCISE 31. Let G be a group of even order. Then G contains an element of order 2.

Proof. Recall for x ∈ G, |x| = 2 iff x 6= 1 and x 2 = 1. Note x 2 = 1 iff x = x −1 . Set

A = { x ∈ G | x = x −1 } and B = { x ∈ G | x 6= x −1 }

Then G = A ∪ B , where this union is disjoint, so |G| = |A| + |B |. Now |B | is even since
the elements of B come in pairs. Since |G| is even by hypothesis, |A| must also be
even. We know 1 ∈ A, hence |A| ≥ 2 and thus there exists x ∈ A, x 6= 1 as desired.

E XERCISE 32. Let G be a group and x ∈ G with |x| = n. Then 1, x, . . . , x n−1 are distinct.

Proof. Suppose there exist 0 ≤ i < j ≤ n − 1 such that x i = x j . Then by cancellation,

x j −i = 1. But 0 < j − i < n, contradicting that n is least in Z+ with x n = 1.

This implies that if |x| = n, G contains at least n distinct elements, so |x| ≤ |G|.
Let H = 〈x〉 (see Exercise 26). If x m ∈ H , then write m = nk + r where 0 ≤ r < n.
We have
x m = x nk+r = (x n )k · x r = 1 · x r = x r
Hence H = {1, x, . . . , x n−1 }. In addition, these elements are distinct by this exercise.
Hence |H | = n. Since x was arbitrary, we have: the order of an element in a group is
the order of the cyclic subgroup generated by that element.

2
E XERCISE 36. Let G = {1, a, b, c} be a group with order 4 and no elements of order 4.
Then the group table for G is uniquely determined. In particular G is abelian.

Proof. Note that 1, a, b, c are all distinct. By Exercise 31, we may assume a 2 = 1. We
compute ab by considering possible cases and using the cancellation laws:

• If ab = 1, then b = a −1 = a—a contradiction.

• if ab = a, then b = 1—a contradiction.
• If ab = b, then a = 1—a contradiction.

Thus we must have ab = c. Similarly ba = c and ac = b = c a. Note that

c 2 = (ba)(ab) = b(a 2 )b = b 2

To compute b 2 , we first note by Exercise 32 and our hypotheses that |b| ≤ 3.

• If b 2 = a, then b 3 = ab = c, so b, b 2 , b 3 6= 1—contradicting |b| ≤ 3.

• If b 2 = b, then b = 1—a contadiction.
• If b 2 = c, then c 2 = c (by the above), so c = 1—a contradiction.

Thus we must have b 2 = 1. This also implies c 2 = 1. Finally this gives bc = a = cb.
The group table for G is thus:
1 a b c
1 1 a b c
a a 1 c b
b b c 1 a
c c b a 1
The symmetry about the diagonal shows that G is abelian. (This also follows from
Exercise 25 and the fact that 12 = a 2 = b 2 = c 2 = 1.)

Section 2

E XERCISE 1( A ). We compute the order of each element in D 6 . Recall

D 6 = { 1, r, r 2 , s, sr, sr 2 }

where r 3 = 1, s 2 = 1, and sr = r s −1 . It is immediate that |1| = 1, |r | = 3, and |s| = 2.

Note that sr 6= 1, but
(sr )2 = s(r s)r = s(sr −1 )r = s 2 = 1
hence |sr | = 2. Similarly, sr 2 6= 1 but

(sr 2 )2 = sr (r s)r r = sr (sr −1 )r r = s(r s)r = s(sr −1 )r = 1

so |sr 2 | = 2.

3
E XERCISE 2. Let x ∈ D 2n and suppose x is not a power of r . Then r x = xr −1 .

Proof. We know we may write x = s k r i for some (unique) 0 ≤ k ≤ 1 and 0 ≤ i ≤ n − 1.

Since x is not a power of r , we must have k = 1, so x = sr i . Thus

r x = r (sr i ) = (r s)r i = (sr −1 )r i = (sr i )r −1 = xr −1

E XERCISE 3. Let x ∈ D 2n and suppose x is not a power of r . Then x has order 2.

Proof. As in the previous exercise, write x = sr i , 0 ≤ i ≤ n − 1. Note x 6= 1, and

x 2 = (sr i )2 = s(r i s)r i = s(sr −i )r i = 1

(where r i s = sr −i follows by induction on i ). Hence |x| = 2.

E XERCISE 7. Let G = 〈 a, b | a 2 = b 2 = (ab)n = 1 〉. If a = s and b = sr where s, r ∈ D 2n ,

then G gives a presentation of D 2n .

Proof. Since s = a and r = s 2 r = ab and r, s generate D 2n , it follows that a, b also

generate D 2n . Let R(G) be the set of relations that hold in G and R(D 2n ) be the set
of relations that hold in D 2n . We claim that R(G) = R(D 2n ).
Indeed, from (ab)n = 1 it follows that r n = 1, from a 2 = 1 it follows that s 2 = 1,
and from a 2 = b 2 it follows that s 2 = (sr )2 = sr sr , so s = r sr , so sr −1 = r s. Since every
relation in D 2n follows from these relations in r, s, we have R(D 2n ) ⊆ R(G).
Working these equations backwards and noting that every relation in G follows
from the relations in a, b (by hypothesis), we have R(G) ⊆ R(D 2n ).

E XERCISE 8. In D 2n , |〈r 〉| = |r | = n (see Exercise 1.32).

E XERCISE 12. Let D be a (regular) dodecahedron in R3 . We count the symmetries

of D in R3 .
Label a face f on D and a vertex v on the face f . Any symmetry of D in R3 must
take f to one of 12 possible face positions, and must take v to one of 5 possible vertex
positions at that face position. Hence there are at most 5 · 12 = 60 symmetries.
On the other hand, since all of these positions are distinct and can be obtained
using (distinct) symmetries, it follows that there are at least 60 symmetries. Hence
there are exactly 60 symmetries.

E XERCISE 14. The set {1, −1} generates Z.

4
E XERCISE 15. Z/nZ = 〈 1 | n · 1 = 1 · · + 1} = 0 〉
| + ·{z
n times

Proof. Recall that Z/nZ = {0, . . . , n − 1}. Thus it is clear that 1 generates Z/nZ.
Suppose r , s ∈ Z/nZ and r = s. We claim that this relation can be derived from
n · 1 = 0. Indeed, write r − s = nk for k ∈ Z. Then we have

r − s = r − s = nk = k(n · 1) = k · 0 = 0

from which r = s follows.

Hence we have a presentation of Z/nZ.

Section 3

E XERCISE 1,3. Define the permutations σ, τ by

σ: 1 7→ 3 2 7→ 4 3 7→ 5 4 7→ 2 5 7→ 1
τ: 1 7→ 5 2 7→ 3 3 7→ 2 4 7→ 4 5 7→ 1

Then we have the following cycle decompositions:

σ = (135)(24) σ2 = (153)
τ = (15)(23) τ2 = 1
στ = (2534) τσ = (1243)
τ2 σ = σ

And we have the following order calculations:

|σ| = 6 |σ2 | = 3 |τ| = 2 |τ2 | = 1 |στ| = |τσ| = 4 |τ2 σ| = 6

E XERCISE 8. Let Ω = {1, 2, . . .}. Then S Ω is infinite.

Proof. S Ω contains the infinite subset S = { (1m) | m = 2, 3, . . . } of transpositions.

E XERCISE 9( B ). Let σ = (12345678). Then by direct computation we have:

σ = (12345678) σ2 = (1357)(2468)
σ3 = (14725836) σ4 = (15)(26)(37)(48)
σ5 = (16385274) σ6 = (1753)(2864)
σ7 = (18765432) σ8 = 1

If i ∈ Z, let k = i mod 8. Then 0 ≤ k < 8 and there exists j ∈ Z such that i = 8 j + k.

Thus we have
σi = σ8 j +k = (σ8 ) j σk = 1 j σk = σk
It follows that σi is an 8-cycle iff i is congruent to one of 1, 3, 5, 7 (mod 8).

5
E XERCISE 10. Let σ = (a 0 a 2 · · · a m−1 ) be an m-cycle. Then |σ| = m.

Proof. Note that σi (a j ) = a j +i mod m (by induction on i ). Since all of the a j are
distinct in the cycle notation, it follows that σi (a j ) 6= a j for 0 < i < m, so σi 6= 1
for 0 < i < m. But σm (a j ) = a j for all j —that is, σm = 1. Hence |σ| = m.

E XERCISE 11. Let σ = (1 2 · · · m), m ∈ Z, m ≥ 1. Then σi is an m-cycle iff (i , m) = 1.

Proof. For notational convenience we work with σ = (0 1 · · · m − 1). Note then

σi (k) = k + i mod m (0 ≤ k ≤ m − 1)

First suppose (i , m) = d > 1. Let j = m/d , l = i /d . Note j < m, and

(σi ) j = σi j = σml = (σm )l = 1l = 1

Thus |σi | < m, so σi is not an m-cycle by Exercise 10.

Conversely, suppose (i , m) = 1. Fix 0 ≤ k ≤ m − 1, and suppose that there exists
0 < j < m such that (σi ) j (k) = k. Then by the first identity above, there must exist
l ∈ Z such that
k + i j = ml + k
that is, i j = ml . But this is a contradiction since j < m and i has no factors from m.
On the other hand,
(σi )m = σi m = σmi = (σm )i = 1i = 1
hence in particular (σi )m (k) = k. Since k was arbitrary, it is clear that σi is an m-
cycle as desired, completing the proof.

E XERCISE 15. Let σ ∈ S n and let m be the least common multiple of the lengths of
the cycles in the cycle decomposition of σ. Then |σ| = m.

Proof. Let σ = σ1 · · · σk be the cycle decomposition of σ. Observe that σ may be

regarded as the product (or composite) in S n of the cycles σi . Since these cycles are
pairwise disjoint by hypothesis, they commute. Hence

σm = (σ1 · · · σk )m = σm m
1 · · · σk

by Exercise 1.24. Now |σi | is just the length of σi , for all 1 ≤ i ≤ k, by Exercise 10
above. Hence since m is a common multiple of these lengths by hypothesis, we
have σm i
= 1 for 1 ≤ i ≤ k. Thus σm = 1.
On the other hand, since m is the least common multiple, if 0 < j < m, then there
j
exists some cycle, say σi , such that j is not a multiple of |σi |. We claim that σi 6= 1.
Indeed, let p = |σi | and write j = pq + r where 0 ≤ r < p. Then
j pq+r p
σi = σi = (σi )q σri = 1q σri = σri
j
If σi = 1, then σri = 1, so r = 0 and j is multiple of |σi |—contradicting our hypothesis.
j
Hence σi 6= 1 as claimed.
Now by the pairwise disjointness of the cycles in σ, this implies that σ 6= 1. Hence
we have shown that m is the least positive integer with σm = 1, that is, |σ| = m.

6
One might conjecture, based on the above exercise, that for any group G and any
arbitrary elements x, y ∈ G with |x| = m and |y| = n, that |x y| = lcm(m, n). But this is
false. For example, let G = S 3 , and set x = (13) and y = (12). Then |x| = 2 and |y| = 2,
so lcm(|x|, |y|) = 2. But x y = (123), so |x y| = 3 6= 2.
The key in the case of cycle decompositions is that the cycles are disjoint, so the
action of one factor cannot ‘alter’ the action of another factor.

E XERCISE 18. We find all numbers n such that S 5 contains an element of order n.
Note that every permutation in S 5 must take one of the following forms:

(a 1 )(a 2 )(a 3 )(a 4 )(a 5 ) (a 1 )(a 2 )(a 3 )(a 4 a 5 ) (a 1 )(a 2 )(a 3 a 4 a 5 )

(a 1 )(a 2 a 3 a 4 a 5 ) (a 1 )(a 2 a 3 )(a 4 a 5 ) (a 1 a 2 )(a 3 a 4 a 5 ) (a 1 a 2 a 3 a 4 a 5 )

where a i , 1 ≤ i ≤ 5, are distinct elements of {1, 2, 3, 4, 5}. Also elements of each of

these forms exist in S 5 . By Exercise 15 above, then, it follows that the orders of the
elements in S 5 are precisely n = 1, 2, 3, 4, 5, 6.
Note that this exercise demonstrates how useful cycle decompositions are. There
are 5! = 120 permutations in S 5 . It would be extremely tedious to solve this problem
by considering the permutations individually.

E XERCISE 20. We find a set of generators and relations for S 3 .

It can be verified that σ = (123) and τ = (12) generate S 3 . Also it can be verified
that the relations σ3 = 1, τ2 = 1, and στ = τσ−1 hold and allow one to write any
element of S 3 in the form τi σ j where 0 ≤ i ≤ 1 and 0 ≤ j ≤ 2. Hence any equality
relation is derivable from these relations, so

S 3 = 〈 σ, τ | σ3 = 1, τ2 = 1, στ = τσ−1 〉

Section 4

E XERCISE 1. |GL 2 (F2 )| = 6

Proof. Let a, b, c, d ∈ F2 . Then

µ ¶
a b a = d = 1 and (b = 0 or c = 0), or
det = ad − bc = 1 ⇐⇒
c d b = c = 1 and (a = 0 or d = 0)

This allows for six possible distinct configurations, hence |GL 2 (F2 )| = 6.

E XERCISE 4. Let n ≥ 0. Then Z/nZ is a field iff n is prime.

Proof. Cases n = 0, 1 are trivial, so we assume n > 1.

Recall that (Z/nZ)× denotes the subset of elements in Z/nZ with multiplicative
inverses. By Proposition 0.4, we know

(Z/nZ)× = { a ∈ Z/nZ | (a, n) = 1 }

where we may choose representatives 0 ≤ a ≤ n − 1. The result follows by observing

that n is prime iff (a, n) = 1 for all 0 < a ≤ n − 1.

7
E XERCISE 5. Let F be a field. Then GL n (F ) is finite iff F is finite.

Proof. If F is finite, then there are only finitely many n × n matrices over F , hence in
particular only finitely many such matrices with inverses, so GL n (F ) is finite.
If F is infinite, then there exist infinitely many matrices in GL n (F ) of the form
 
a
 ..  = a · I n×n

 .
a

with a ∈ F .

E XERCISE 8. Let F be a field and n ≥ 2. Then GL n (F ) is nonabelian.

Proof. Let 0, 1 ∈ F denote the additive and multiplicative identities in F , respectively.

Recall that 0 6= 1 by the field axioms. Set
   
1 ··· 1 1
.. ..  .
A= and B = ..
  
. . 
1 1

(where a blank entry indicates a 0). Then

AB (1,1) = 1 · 0 + · · · + 1 · 0 + 1 · 1 = 1 6= 0 = 0 · 1 + 0 · 0 + · · · + 0 · 0 + 1 · 0 = B A (1,1)

So AB 6= B A and hence GL n (F ) is nonabelian.

Section 6

E XERCISE 1. Let G, H be groups and ϕ : G → H be a homomorphism. Then for all

x ∈ G and for all n ∈ Z, ϕ(x n ) = ϕ(x)n .

Proof. Fix x ∈ G. Note that

ϕ(1G ) = ϕ(1G · 1G ) = ϕ(1G ) · ϕ(1G )

hence ϕ(1G ) = 1H by cancellation in H . This establishes the result for n = 0.

Note in addition that

ϕ(x) · ϕ(x −1 ) = ϕ(xx −1 ) = ϕ(1G ) = 1H

Thus ϕ(x −1 ) = ϕ(x)−1 by uniqueness of inverses, so the result holds for n = −1.
For n > 0, proceed by induction. If the result holds for n ≥ 0, then

ϕ(x n+1 ) = ϕ(x n · x)

= ϕ(x n ) · ϕ(x)
= ϕ(x)n · ϕ(x)
= ϕ(x)n+1

8
And hence the result holds for n + 1. Finally, if the result holds for n < 0, then

ϕ(x n−1 ) = ϕ(x n · x −1 )

= ϕ(x n ) · ϕ(x −1 )
= ϕ(x)n · ϕ(x)−1
= ϕ(x)n−1

Hence the result holds for n − 1.

This establishes the result for all n ∈ Z.

E XERCISE 2. Let ϕ : G → H be an isomorhism. Then |ϕ(x)| = |x| for all x ∈ G. In

addition, G and H have the same number of elements of each order n, for all n > 0.

Proof. Fix x ∈ G and let n = |x|. Then

ϕ(x)n = ϕ(x n ) = ϕ(1G ) = 1H

by Exercise 1. On the other hand, if 0 < m < n, then x m 6= 1G , and hence ϕ(x)n 6= 1H
by injectivity of ϕ. So |ϕ(x)| = n as desired.
For each n > 0, let

X n = { x ∈ G | |x| = n } and Yn = { y ∈ H | |y| = n }

Note ϕ| X n is an injection from X n into Yn . Moreover ϕ| X n is a surjection since

ϕ is a surjection. Thus |X n | = |Yn |. This establishes the second part of the result.

E XERCISE 3. Let ϕ : G → H be an isomorphism. Then G is abelian iff H is abelian.

Proof. Suppose G is abelian. Let a 0 , b 0 ∈ H and set a = ϕ−1 (a 0 ) and b = ϕ−1 (b 0 ). Then

a 0 b 0 = ϕ(a)ϕ(b) = ϕ(ab) = ϕ(ba) = ϕ(b)ϕ(a) = b 0 a 0

Hence H is abelian.
Interchanging the roles of G and H establishes the converse.

E XERCISE 4. R× ∼
6 C×
=

Proof. Recall that i ∈ C× and |i | = 4. We claim that R× has no elements of order 4.

Indeed, if r ∈ R and r 4 = 1, then (r 2 )2 = 12 , so r 2 = 1 by the uniqueness of non-
negative (square) roots in R. Hence |r | ≤ 2.
The result follows from Exercise 4.

E XERCISE 5. R ∼
6 Q
=

Proof. Immediate since |R| > |Q|.

9
E XERCISE 6. Q ∼
6 Z
=

Proof. Suppose ϕ : Q ∼
= Z. Set m = ϕ(1) and fix n > 0. Then

m = ϕ(1) = ϕ(n · (1/n)) = n · ϕ(1/n)

hence ϕ(1/n) = m/n ∈ Z. But this is impossible as n → ∞.

E XERCISE 9. D 24 ∼
6 S4
=

Proof. We know that D 24 has an element of order 12. However, by looking at cycle
decompositions (cf. Exercise 3.18), we see that S 4 only has elements of orders m =
1, 2, 3, 4. Thus the result follows from Exercise 4.

E XERCISE 10. Let ∆, Ω be sets and suppose |∆| = |Ω|. Then S ∆ ∼

= SΩ.

Proof. Let π : Ω → ∆ be a bijection. Define Φ : S ∆ → S Ω by

Φ : σ 7→ π−1 ◦ σ ◦ π

Note that Φ is well defined since if σ ∈ S ∆ , then Φ(σ) ∈ S Ω (clearly Φ(σ) : Ω → Ω, and
Φ(σ) is a bijection since the composite of bijections is a bijection).
Note that Φ is a homomorphism since if σ, τ ∈ S ∆ , then

Φ(στ) = π−1 (στ)π = (π−1 σπ)(π−1 τπ) = Φ(σ)Φ(τ)

Also Φ is injective by cancellation. Finally, if θ ∈ S Ω , set σ = πθπ−1 . Then σ ∈ S ∆ and

Φ(σ) = θ, so Φ is surjective.

Note how conjugation captures a type of ‘translation’ process.

For the next problem we need a definition: if ϕ : G → H is a homomorphism,
define
ker ϕ = { x ∈ G | ϕ(x) = 1H }
called the kernel of ϕ. It is immediate that ker ϕ ≤ G.

E XERCISE 14. Let ϕ : G → H be a homomorphism. Then ϕ is injective iff ker ϕ = 1.

Proof. Suppose ker ϕ = 1. If ϕ(a) = ϕ(b), then

ϕ(ab −1 ) = ϕ(a)ϕ(b)−1 = 1

hence ab −1 ∈ ker ϕ. By hypothesis, then, ab −1 = 1, so a = b. Thus ϕ is injective.

Conversely, suppose ϕ is injective. We know ϕ(1) = 1, hence if a ∈ ker ϕ, then we
must have a = 1. Thus ker ϕ = 1.

10
E XERCISE 19. Let G = { z ∈ C | (∃n > 0)[ z n = 1 ] } and fix k > 1. Then the map z 7→ z k is
a surjective homomorphism from G to G, but is not an isomorphism.

Proof. Indeed, (z 1 z 2 )k = z 1k z 2k . If z ∈ G with z n = 1, let w ∈ C be any k-th root of z.

Then w k = z, and since z n = (w k )n = w kn = 1, we have w ∈ G.
But the map is not injective since the kernel consists of the k-th roots of unity,
and there are k of these.

E XERCISE 25. Let n > 0 and set θ = 2π/n. Recall from linear algebra that the matrices

cos θ − sin θ
µ ¶ µ ¶
0 1
R= S=
sin θ cos θ 1 0

in GL 2 (R) represent in R2 a counterclockwise rotation about the origin by θ radians,

and a reflection about the line y = x, respectively. Let r, s generate D 2n as usual.
We claim that ϕ : D 2n → GL 2 (R) defined on the generators by r 7→ R and s 7→ S
extends uniquely to an injective homomorphism.

Proof. Indeed, we define ϕ on D 2n by

r k1 s k2 · · · r km s km+1 7→ R k1 S k2 · · · R km S km+1

This map is defined on all of D 2n since r, s generate D 2n . Moreover it is well defined

and injective since R, S satisfy precisely the same relations in GL 2 (R) as r, s satisfy
in D 2n . This map clearly extends the definition given on generators above, and it is
the unique extension since any homomorphism mapping the generators that way
must agree with it.

Note that the map is not surjective since there are more elements in GL 2 (R) than
those generated by R, S. We note that the map ϕ provides a matrix representation of
the dihedral group (cf. permutation representations).

Section 7

E XERCISE 3. The additive group R acts on the plane R × R by r · (x, y) = (x + r y, y).

Proof. For x, y ∈ R, 0 · (x, y) = (x + 0y, y) = (x, y), so the identity property holds. Also,
if r, s ∈ R,

(r + s) · (x, y) = (x + (r + s)y, y)
= (x + r y + s y, y)
= (x + s y + r y, y)
= r · (x + s y, y) = r · (s · (x, y))

Hence associativity holds as well.

Note that this action is just a horizontal shear. The geometric fact that horizontal
shears accumulate additively is reflected in the fact that the action is given by R. (To
compare, see that the action is not given by R× , for example.)

11
E XERCISE 8( A ). Let A be a nonempty set. Fix k ∈ Z, 0 < k ≤ |A|, and let A be the set
of all subsets of A of cardinality k. Then S A acts on A by

σ · {a 1 , . . . , a k } 7→ {σ(a 1 ), . . . , σ(a k )}

Proof. The identity and associativity properties hold trivially.

We include this problem because it illustrates how a group of permutations of one

type of object (in this case, elements of a set A) may naturally give rise to a group of
permutations of another type of object. This is also seen, for example, in the regular
action of S A on itself.

E XERCISE 11. Let D 8 act on the vertices of a square. Label the vertices consecutively
1, 2, 3, 4, starting from a fixed vertex and moving in a clockwise manner. We write
cycle decompositions of the elements in S 4 corresponding to the elements of D 8
under this action.

Proof. We have the following representation:

1 7→ 1
r 7→ (1234)
r 2 7→ (13)(24)
r 3 7→ (1423)
s 7→ (24)
sr 7→ (24)(1234) = (14)(23)
sr 2 7→ (24)(13)(24) = (13)
sr 3 7→ (24)(1423) = (12)(34)

Note that this map is injective, hence we have obtained a faithful representation
of D 8 in S 4 . In other words, D 8 is isomorphic to a subgroup of S 4 (though note this
must be a proper subgroup, since |D 8 | = 8 and |S 4 | = 24). Compare this with the
faithful (matrix) representation obtained in Exercise 6.25.

E XERCISE 12. Let n ∈ Z+ be even, and label the vertices of a regular n-gon as above
with labels from V = {1, . . . , n}. Set

P = { {a, b} | a, b ∈ V and a, b label opposite vertices }

Then D 2n acts on P .

Proof. For convenience we identify V with the set of vertices of the n-gon, and
P with the set of pairs of opposite vertices.

12
 Let φ : D 2n → S V be the representation of D 2n given by the usual action of D 2n
on V . Define ψ : D 2n → S P as follows: for α ∈ D 2n , set

ψ(α) : P → P
{a, b} 7→ {φ(α)(a), φ(α)(b)}

Note that ψ(α) is a well defined permutation of P , hence ψ is well defined. Also ψ is
a homomorphism since φ is. Thus ψ defines an action of D 2n on P .

We calculate the kernel of this action by cases:

1. If n = 2, then direct computation reveals that the action is trivial (that is, the
kernel is all of D 4 ).

2. If n = 4, then direct computation reveals that

ker ψ = {1, r 2 , s, sr 2 }

where r is the usual clockwise rotation through θ = π/2, and s the reflection in
the line passing through opposing vertices 1 and 3.

3. If n > 4, write n = 2m. Then we claim ker ψ = {1, r m }, where r is a clockwise

rotation through θ = 2π/n.
Indeed, clearly {1, r m } ⊆ ker ψ. For the converse, first note that there are no
other orientation preserving symmetries (powers of r ) in ker ψ. Suppose now,
towards a contradiction, that α ∈ ker ψ is orientation reversing. We may write
α = r k s, where s is the reflection in the line passing through vertex 1 and its
opposing vertex. Let p 1 , p 2 , p n ∈ P , where 1 ∈ p 1 , 2 ∈ p 2 , and n ∈ p n . Note
that (the action of) s preserves p 1 , but interchanges p 2 and p n . Thus, since
α ∈ ker ψ, r k must interchange p 2 and p n again, while fixing p 1 . But this is
impossible. Hence there are no orientation reversing symmetries in ker ψ. It
follows that ker ψ = {1, r m } as claimed.

E XERCISE 16. Let G be a group. Then G acts on itself by conjugation, where

g · x = g xg −1

for all g , x ∈ G.

Proof. Clearly the identity property holds. If g 0 , g 2 , x ∈ G, then

(g 0 g 2 ) · x = (g 0 g 2 )x(g 0 g 2 )−1
= (g 0 g 2 )x(g 2−1 g 0−1 )
= g 0 (g 2 xg 2−1 )g 0−1
= g 0 (g 2 · x)g 0−1 = g 0 · (g 2 · x)

Hence associativity holds as well.

13
E XERCISE 17. Let G be a group and fix g ∈ G. Then the operation of conjugation by g
is an automorphism of G.

Proof. Indeed, let x, y ∈ G. Then

g (x y)g −1 = g (x(g −1 g )y)g −1 = (g xg −1 )(g y g −1 )

Hence conjugation by g is a homomorphism. We know that it is a bijection by the

previous exercise. Hence it is an automorphism as desired.

The previous two exercises show that conjugation in a group G is a very special group
action, since from each group element we obtain not merely a permutation but an
automorphism of G.
As an example, consider G = GL 2 (R), which for convenience we identify with the
group of invertible linear transformations on R2 . Each element g ∈ G gives rise to
a transformation of R2 . Now if x ∈ G, and we view x as an operation on the pre-
transformed space, then the transformation naturally corresponding to x in the g -
transformed space is simply g xg −1 . Thus the operation of g on all of G produces
a ‘translated version’ of G, where every x ∈ G is ‘relativized’ to g . This ‘translated
version’ of G is structurally identical to G.

E XERCISE 18. Let H be a group acting on a set A. Define a relation ∼ on A by

a ∼ b ⇐⇒ (∃h ∈ H )(a = hb)

Then ∼ is an equivalence relation on A.

Proof. Let a, b, c ∈ A be arbitrary. Then a ∼ a since a = 1 · a by the identity property

of the action. If a ∼ b, choose h ∈ H with a = hb. We have

b = 1 · b = (h −1 h)b = h −1 (hb) = h −1 a

hence b ∼ a. Finally, if a ∼ b and b ∼ c, choose h, g ∈ H with a = hb and b = g c. Then

a = hb = h(g c) = (hg )c

hence a ∼ c. Since a, b, c were arbitrary, this shows that reflexivity, symmetry, and
transitivity hold for ∼, hence ∼ is an equivalence relation on A.

The equivalence class of a is called the orbit of a under the action of H .

E XERCISE 19 (L AGRANGE ). Let G be a finite group and H a subgroup of G. Then

|H | divides |G|.

Proof. Let H act on G by the left regular action. By the previous exercise, and the
assumption that G is finite, we may partition G into finitely many orbits under H :

G = O(x 1 ) ∪ · · · ∪ O(x n )

But note that O(x i ) is merely the image of H under the permutation (in G) of right
multiplication by x i . Hence |O(x i )| = |H | for all 1 ≤ i ≤ n. Since G is partitioned,
|G| = n · |H |, so the result holds.

14
E XERCISE 23. Let G be the group of rotations of a cube, and let G act on the set
A = {p 1 , p 2 , p 3 } of three pairs of opposite faces of the cube. Then the kernel of this
action is the subgroup H of G generated by {1, r 1 , r 2 , r 3 }, where r i is the rotation by π
radians about the line passing through the centers of the two faces in p i .

Proof. It is immediate that the kernel contains all of H , since it contains {1, r 1 , r 2 , r 3 }.
To prove the converse, let ρ be any rotation in the kernel, and consider the effect
of ρ on p 1 . Either ρ swaps the two faces in p 1 or it does not. In the non-swapping
case, either ρ = 1 or ρ = r 1 . In the swapping case, it is easy to see that either ρ = r 2 or
ρ = r 3 . Hence H contains the kernel.
This shows that H is the kernel.

In fact we see from the above proof that H = {1, r 1 , r 2 , r 3 }. In addition we see that the
action’s representation of G is not faithful.
Note that the above proof could be simplified by just arguing directly that any
element in the kernel is one of 1, r 1 , r 2 , or r 3 . We leave the above proof as it is to
illustrate the technique of using a subgroup generated by particular elements; in
more complex cases, it might not be as easy to obtain an explicit list of the elements
in the kernel.

Chapter 2
Section 1

E XERCISE 1.

(b) Let G = C× and H = { z ∈ G | |z| = 1 }. Then H ≤ G. Proof: 1 ∈ H . If z, w ∈ H ,

then |zw| = |z||w| = 1 · 1 = 1, so zw ∈ H . Similarly |1/z| = 1/|z| = 1, so 1/z ∈ H .

(e) Let G = R× and H = { r ∈ G | r 2 ∈ Q }. Then H ≤ G. Proof: 1 ∈ H . If r, s ∈ H , then

r 2 = m 0 /n 1 and s 2 = m 00 /n 2 for some m i , n i ∈ Z − {0}. But then
µ 0 ¶ µ 00 ¶
m m m 0 m 00
(r s)2 = r 2 · s 2 = = ∈Q
n1 n2 n1 n2

hence r s ∈ H . Similarly (1/r )2 = 1/r 2 = n 1 /m 0 ∈ Q, so 1/r ∈ H .

E XERCISE 2.

(a) Fix n ≥ 3, set G = S n and let S be the set of 2-cycles in G. Then S 6≤ G. Proof:
Note that (12), (13) ∈ S, but (13)(12) = (123) 6∈ S.
p p
(e) Let G = R and let S = { r ∈ G | r 2 ∈ Q }. Then S 6≤ G. Proof: Note 2, 3 ∈ S, but
p p p p p
( 2 + 3)2 = 2 + 2 2 3 + 3 = 5 + 6
p p p
which is not rational since 6 is not rational. Hence ( 2 + 3) 6∈ S.

15
E XERCISE 4. Let G = Z and let N be the subset of nonnegative integers. Then N is
infinite and closed under the group operation, but N is not a subgroup since it does
not contain additive inverses.
Alternately let G = Q× . Then Z is an infinite subset and is closed under the group
operation, but it is not a subgroup since it does not contain multiplicative inverses.

E XERCISE 5. Let G be a group suppose n = |G| > 2. Then G does not have a subgroup
of order n − 1.

Proof. Suppose H ≤ G and |H | = n − 1. Since |H | > 1, we may choose h ∈ H , h 6= 1.

Let g be the unique element of G − H . Now hg 6= g , hence hg ∈ H . But then since
h ∈ H , we have h −1 ∈ H and
h −1 (hg ) = (h −1 h)g = 1 · g = g ∈ H
—contradicting the fact that g 6∈ H .

E XERCISE 6. Let G be an abelian group and define

H = { g ∈ G | |g | < ∞ }
Then H is a subgroup of G (called the torsion subgroup of G).

Proof. Clearly 1 ∈ H . Suppose x, y ∈ H , and let m = |x| and n = |y|. Then

(x y)mn = x mn y mn = (x m )n (y n )m = 1n · 1m = 1 · 1 = 1
Thus |x y| < ∞, so x y ∈ H . Similarly (x −1 )m = (x m )−1 = 1−1 = 1, so x −1 ∈ H .

E XERCISE 7. Fix n ∈ Z, n > 1 and let G = Z × (Z/nZ). Then G is an abelian group, and
it is clear that the torsion subgroup of G is { (0, a) | a ∈ Z/nZ }.
The subset of G consisting of the elements of infinite order (together with 0) is
not a subgroup. Indeed, (−1, 0) and (1, 1) have infinite order, but (−1, 0)+(1, 1) = (0, 1)
has finite order (and is not 0).

E XERCISE 8. Let H , K ≤ G. Then H ∪ K ≤ G iff H ⊆ K or K ⊆ H .

Proof. One direction is trivial.

Suppose H 6⊆ K and K 6⊆ H . Choose h ∈ H − K and k ∈ K − H . Then h, k ∈
H ∪ K , but hk 6∈ H ∪ K . Indeed, if hk ∈ H , then h −1 (hk) = (h −1 h)k = k ∈ H —a
contradiction—and if hk ∈ K , then (hk)k −1 = h(kk −1 ) = h ∈ K —a contradiction.
Thus H ∪ K is not a subgroup of G.

E XERCISE 9. Let G = GL n (F ), F an arbitrary field. Define

SL n (F ) = { A ∈ GL n (F ) | det(A) = 1 }
Then SL n (F ) ≤ GL n (F ) (called the special linear group).

Proof. Clearly 1 ∈ SL n (F ). If A, B ∈ SL n (F ), then AB ∈ SL n (F ) since

det(AB ) = det(A) det(B ) = 1 · 1 = 1
−1
Similarly A ∈ SL n (F ) since det(A −1 ) = 1.

16
E XERCISE 10( B ). Let G be a group and H be an arbitrary nonempty collection of
subgroups of G. Then H ≤ G.
T

Proof. Note that H ⊆ G since H is nonempty. Now 1 ∈ H since 1 ∈ H for all

T T

H ∈ H . Suppose x, y ∈ H . If H ∈ H is arbitrary, then x, y ∈ H , so x y ∈ H since

T

H ≤ G. Since H was arbitrary, x y ∈ H . Similarly x −1 ∈ H . Thus H ≤ G.

T T T

E XERCISE 13. Let 0 < H ≤ Q and suppose 1/x ∈ H for all nonzero x ∈ H . Then H = Q.

Proof. Fix x ∈ H , x 6= 0. Write x = m/n where m, n ∈ Z, m 6= 0 and n > 0. Note that

³m ´
m=n· = n · x = |x + ·{z
· · + x} ∈ H
n
n times

Hence m ∈ H , and since m 6= 0, 1/m ∈ H . But then, as above, m · (1/m) = 1 ∈ H .

Now let y ∈ Q be arbitrary. Write y = m 0 /n 0 , with m 0 , n 0 ∈ Z and n 0 > 0. We have
n = n 0 · 1 ∈ H since 1 ∈ H . Since n 0 > 0, 1/n 0 ∈ H . Finally, m 0 /n 0 = m 0 · (1/n 0 ) ∈ H .
0

Thus, since y was arbitrary, H = Q.

This result shows, in particular, that Q is the smallest additive subgroup containing Z
and closed under multiplicative inverses.

E XERCISE 15. Let G be a group and H be a nonempty chain of subgroups of G. Then

H ≤ G.
S

Proof. Fix H ∈ H . Then 1 ∈ H ⊆ H . Suppose x, y ∈ H . Then x ∈ H and y ∈ K for

S S

some H , K ∈ H . Since H is a chain, either H ⊆ K or K ⊆ H , so H ∪ K is a subgroup

equal to one of them (cf. Exercise 8). Now x, y ∈ H ∪ K , so x y ∈ H ∪ K ⊆ H . Also
S

x −1 ∈ H ⊆ H . It follows that H ≤ G.
S S

Section 2

E XERCISE 6. Let H ≤ G. Then H ≤ NG (H ), and H ≤ CG (H ) iff H is abelian.

Proof. For the first claim, let h ∈ H . Then hH h −1 ⊆ H . If h 0 ∈ H , then h −1 h 0 h ∈ H , so

h 0 = h(h −1 h 0 h)h −1 ∈ hH h −1 . Thus hH h −1 = H . Since h was arbitrary, H ≤ NG (H ).
The second claim is immediate from definitions.

E XERCISE 7. Let n ∈ Z, n ≥ 3. If n is odd, Z (D 2n ) = 1, and if n is even, Z (D 2n ) = {1, r k },

where n = 2k.

Proof. Note that Z (D 2n ) = C (r ) ∩ C (s) since r, s generate D 2n . Now we know that

C (r ) = 〈r 〉 = {1, r, . . . , r n−1 }. We also know {1, s} ≤ C (s). We claim that r i ∈ C (s) for
some 0 < i < n iff n is even, and in this case n = 2i .
Indeed, suppose r i ∈ C (s) for 0 < i < n. Then sr i = r i s = sr −i , so by cancellation
i
r = r −i , or r 2i = 1. It follows that n|2i . Now if n is odd, then n|i —contradicting that
0 < i < n. Hence n must be even. In addition, since 2i < 2n, we must have 2i = n.
The converse claim is trivial.
Now if n is odd, then C (s) = {1, s}, hence we must have Z (D 2n ) = 1. If n is even,
then C (s) = { 1, r k , s, sr k }, where n = 2k. Hence Z (D 2n ) = {1, r k }.

17
In the above proof, we see an instance of the general result that if G is a group, S ⊆ G,
T
and G = 〈S〉, then Z (G) = {CG (s) | s ∈ S }. In particular, |Z (G)| divides |CG (s)| for all
s ∈ S (by Lagrange’s Theorem).
The following result is also useful in the computation of centralizers: if G is a
group and g ∈ G, then CG (g ) = CG (〈g 〉). In particular, if |g | = n, and (m, n) = 1, then
CG (g m ) = CG (g ) since 〈g m 〉 = 〈g 〉.

E XERCISE 13. Fix n ∈ Z, n > 0 and let R be the set (ring) of all polynomials in x 1 , . . . , x n
over Z. For each σ ∈ S n , define the map σ : R → R by

σ : p(x 1 , . . . , x n ) 7→ p(x σ(1) , . . . , x σ(n) )

This mapping defines a (left) action of S n on R.

Proof. The identity and associativity properties are immediate.

From this exercise we see that the permutations in S n naturally induce permutations
on polynomials in n variables.

Section 3

N OTE . Let G = 〈x〉. Then

(a) If |x| = n, 〈x a 〉 ≤ 〈x b 〉 iff (b, n)|(a, n).

Proof. If 〈x a 〉 ≤ 〈x b 〉, then |x a | divides |x b | by Lagrange and Proposition 2. By

Proposition 5, this means
n n
·k =
(a, n) (b, n)
for some k ∈ Z. In other words, (b, n) · k = (a, n), so (b, n)|(a, n) as desired.
Conversely, if (a, n) = k · (b, n) for some k ∈ Z, then

x (a,n) = x k·(b,n) = (x (b,n) )k ∈ 〈x (b,n) 〉

Hence 〈x (a,n) 〉 ≤ 〈x (b,n) 〉. But 〈x a 〉 = 〈x (a,n) 〉 and 〈x b 〉 = 〈x (b,n) 〉 by Theorem 7(3).

This completes the proof.

(b) If |x| = ∞, 〈x a 〉 ≤ 〈x b 〉 iff b|a.

Proof. If x a ∈ 〈x b 〉, then x a = (x b )k = x bk for some k ∈ Z. This gives x a−bk = 1,

so we must have a − bk = 0 since |x| = ∞. Thus a = bk, or b|a as desired.
The converse is immediate.

It follows from these results that for G = 〈x〉, if |G| = n, then 〈x b 〉 = G iff (b, n) = 1;
and if |G| = ∞, then 〈x b 〉 = G iff b = ±1 (cf. Proposition 6).
We also obtain the following specific case (cf. p. 59): if n ≥ 1 and 1 ≤ a, b ≤ n,
then 〈a〉 ≤ 〈b〉 in Z/nZ iff (b, n)|(a, n).

18
E XERCISE 1. We determine the subgroup structure (lattice) of Z45 = 〈x〉.
Note that 45 = 3 · 3 · 5, with positive divisors 1, 3, 5, 9, 15, 45. Hence by Theorem 7,
the subgroups of Z45 are

〈1〉 〈x 15 〉 〈x 9 〉 〈x 5 〉 〈x 3 〉 〈x〉

By the result above, we see that 〈x a 〉 ≤ 〈x b 〉 iff (b, 45)|(a, 45) for 1 ≤ a, b ≤ 45.

E XERCISE 5. Let G = Z/49000Z. Note that 49000 = 23 · 53 · 72 , hence by Proposition 6,

the number of generators in G is given by

ϕ(49000) = ϕ(23 ) · ϕ(53 ) · ϕ(72 )

= 22 (1) · 52 (4) · 7(6)
= 4 · 100 · 42
= 16800

E XERCISE 12. The following groups are noncyclic: Z2 × Z2 , Z2 × Z, and Z × Z.

Proof. Note that each element in Z2 ×Z2 has order at most 2, hence the group cannot
be cyclic (for this requires an element of order 4).
If (a, b) generates Z2 × Z, then we must have b = ±1 by Proposition 6. Clearly
we cannot have a = 1, for 〈(1, b)〉 = 1 × Z < Z2 × Z. But we also cannot have a = x,
because for example (1, 1) 6∈ 〈(x, b)〉. Since these possibilities are exhaustive, there is
no generator for the group.
Let (a, b) ∈ Z × Z. Define
(
(1, −1) if a and b have the same sign
x=
(1, 1) if not

Then x 6∈ 〈(a, b)〉. Since (a, b) was arbitrary, there is no generator for the group.

E XERCISE 13. Note that Z × Z2 is not isomorphic to Z since Z is cyclic but Z × Z2 is

not (cf. Exercise 12).
We claim that Q × Z2 is not isomorphic to Q. Indeed, if ϕ : Q × Z2 → Q is an
injective homomorphism, then ϕ(0, x) = q 6= 0. But then

ϕ((0, x) · (0, x)) = ϕ(0, 1) = 0 6= q + q = ϕ(0, x) + ϕ(0, x)

—a contradiction.

E XERCISE 15. Note that Q × Q is not cyclic by Theorem 7 since 1 × Q ≤ Q × Q but

1×Q ∼= Q is not cyclic (see Exercise 4.14(d) below).

19
E XERCISE 16. Let G be a group and suppose x, y ∈ G commute. Set m = |x| and
n = |y|. Then |x y| divides lcm(m, n).

Proof. Indeed, write α = lcm(m, n). Since m|α and n|α, x α = 1 and y α = 1. Then
since x and y commute, (x y)α = x α · y α = 1. The result follows from Proposition 3.

Note that |x y| need not equal lcm(m, n). Indeed, let G = S 3 and set x = y = (12).
Then m = n = 2, so lcm(m, n) = 2, but x y = 1 so |x y| = 1 < 2.
Also the result need not hold if x, y do not commute. Again in S 3 , set x = (12) and
y = (13). Then lcm(m, n) = 2. But x y = (132), so |x y| = 3, which does not divide 2.

E XERCISE 23. The group (Z/2n Z)× is noncyclic for n ≥ 3.

Proof. Fix n ≥ 3 and let G = (Z/2n Z)× . We claim that x = 2n−1 −1 and y = 2n−1 +1 are
two distinct elements of order 2 in G, from which it follows that G cannot be cyclic
by Theorem 7(3).
Note that 0 < x < y < 2n . Thus if x = 1 (in G), then 2n−1 − 1 = 1 (in Z), which
implies 2n−2 = 1—a contradiction since n ≥ 3. On the other hand,

x 2 = (2n−1 − 1)2 = (2n−1 )2 − 2 · 2n−1 + 1 = (2n )2 · 2−2 − 2n + 1 = 1

Hence |x| = 2. Similarly y 6= 1 but since y = −x, y 2 = (−x)2 = x 2 = 1, so |y| = 2.

It is easy to check that x 6= y, hence x and y are two distinct elements of order 2,
establishing our claim.

E XERCISE 25. Let G be a cyclic group of order n and let k be relatively prime to n.
Then the map ϕ : g 7→ g k is an automorphism of G.

Proof. Write G = 〈x〉. Then ϕ is a homomorphism since

ϕ(x a · x b ) = ϕ(x a+b )

= (x a+b )k
= x (a+b)k
= x ak · x bk
= ϕ(x a ) · ϕ(x b )

Now 〈x k 〉 = G since (k, n) = 1 by Proposition 6. Hence for any x a ∈ G, there exists m

such that x a = (x k )m = x km = (x m )k , so ϕ(x m ) = x a . Therefore ϕ is a surjective
homomorphism. But since G is finite, this means ϕ is an automorphism.

If G is an arbitrary group of order n and (k, n) = 1, then the map ϕ : g → g k may

not be a homomorphism. For example, let G = D 6 (so n = 6) and let k = 5. Then
(k, n) = 1, but
ϕ(r s) = (r s)5 = r s 6= r 2 s = r 5 · s 5 = ϕ(r ) · ϕ(s)

20
However, in general the map is still surjective. Indeed, for G arbitrary and g ∈ G, let
H = 〈g 〉 and let m = |g |. Then by Lagrange’s Theorem, m|n, hence (k, m) = 1. Note
that ϕ|H : H → H . Hence by the above exercise there exists h ∈ H such that ϕ(h) = g .
On the other hand, if (k, n) > 1, then by Cauchy’s Theorem (Section 3.2) there
exists x ∈ G with |x| = d > 1 and d |k. Then x 6= 1 but x k = 1, so the the map g 7→ g k
is not injective, and hence not surjective since G is finite. Thus we obtain: for G of
order n, the map g 7→ g k is surjective iff (k, n) = 1.

E XERCISE 26. Let Zn be a cyclic group of order n. For a ∈ Z, define

σa : Z n → Z n by σa (x) = x a (x ∈ Zn )

(a) σa is an automorphism of Zn iff (a, n) = 1. (Proof: previous exercise).

(b) σa = σb iff a ≡ b (mod n).

Proof. If a ≡ b, then a = b + nk for some k ∈ Z, hence

σa (x) = x a = x b+nk = x b · (x n )k = x b = σb (x)

for all x ∈ Zn , that is, σa = σb . Conversely, if σa = σb and Zn = 〈y〉, then y a =

y b , so y a−b = 1, so n|(a − b), so a ≡ b.

(c) Every automorphism of Zn is of the form σa for some a ∈ Z.

Proof. Let σ be an automorphism of Zn and write Zn = 〈y〉. Then σ(y) = y a

for some a ∈ Z. If x ∈ Zn , then x = y b for some b ∈ Z , hence

σ(x) = σ(y b ) = σ(y)b = (y a )b = y ab = (y b )a

Thus σ = σa .

(d) (Z/nZ)× ∼
= Aut(Zn )

Proof. We claim the map a 7→ σa witnesses the isomorphism. Indeed, by

part (b) this map is well defined and injective. By parts (a)–(c), the map is
surjective. And the map satisfies the homomorphism property since

σab (x) = x ab = (x b )a = σa ◦ σb (x)

for all x ∈ Zn , hence σab = σa ◦ σb .

Note that this result shows that Aut(Zn ) is an abelian group of order ϕ(n). By
Theorem 4, this result actually gives the structure of the automorphism group
for any finite cyclic group of order n.

Section 4

21
E XERCISE 3. Suppose H ≤ G is abelian. Then 〈H , Z (G)〉 is abelian.

Proof. Let x, y ∈ 〈H , Z (G)〉. Then by Proposition 9, x and y are both finite products
of elements all of which commute with each other. Hence x y = y x.

Note this result does not hold in general for the centralizer of an abelian subgroup.
Let G = S 5 and H = 〈(12)〉. Then H is abelian, and (34), (45) ⊆ C (H ), but 〈H ,C (H )〉 is
nonabelian since (34)(45) = (345) 6= (354) = (45)(34).

E XERCISE 5. Suppose x, y ∈ S 3 with x 6= y and |x| = 2 = |y|. Then 〈x, y〉 = S 3 .

Proof. Note that {1, x, y, x y} ⊆ 〈x, y〉. It is easily verified using cancellation that these
elements are distinct, hence |〈x, y〉| ≥ 4. But then by Lagrange’s Theorem, we must
have |〈x, y〉| = 6, that is, 〈x, y〉 = S 3 .

E XERCISE 6. Let x = (12) and y = (12)(34) (in S 4 ). Then 〈x, y〉 is noncyclic of order 4.

Proof. It is easily verified that 〈x, y〉 = {1, x, y, x y}, where these elements are distinct.
Hence |〈x, y〉| = 4. But there are no elements in this group of order 4, hence it is
noncyclic (and isomorphic to the Klein Four group by Exercise 1.1.36).

E XERCISE 13. Let G = (Q+ )× . Then G = 〈 1/p | p prime 〉.

Proof. Write S = { 1/p | p prime } and H = 〈S〉. We require H = G.

Let q ∈ G. Then q = m/n for some m, n ∈ Z+ . By prime factorization, write
n = p 1 · · · p j . Note that 1/p 1 , . . . , 1/p j ∈ S, hence the product
³ 1 ´ ³ 1 ´ 1 1
··· = = ∈H
p1 pj p1 · · · p j n
Similarly 1/m ∈ H , and hence 1/(1/m) = m ∈ H . But then m · (1/n) = m/n = q ∈ H .
Since q was arbitrary, H = G.

E XERCISE 14.
(c) Let H ≤ Q be finitely generated. Then H is cyclic.

Proof. Write H = 〈A〉 where A = { m 0 /n 1 , . . . , m k /n k }. Set q = 1/(n 1 · · · n k ). Then

A ⊆ 〈q〉, since
mi
= (m i n 1 · · · n i −1 n i +1 · · · n k ) · q
ni
Hence H ≤ 〈q〉. By Theorem 2.3.7, H is cyclic.

(d) Q is not finitely generated.

Proof. This follows from (c) and the fact that Q is not cyclic: for q = m/n > 0
arbitrary, note that
m m
−kq < 0 < < ≤ kq (k ∈ Z, k > 0)
n +1 n
Hence 〈q〉 < Q. Since q was arbitrary, Q is noncyclic.

22
E XERCISE 16.

(a) Let H be a proper subgroup of a finite group G. Then there exists a maximal
subgroup of G containing H .

Proof. Since G is finitely generated, this follows from the proof of Exercise 17
by setting H0 = H .

(b) Let G = D 2n = 〈 r, s | r n = s 2 = 1, r s = sr −1 〉 and set H = 〈r 〉. Then H is maximal.

Proof. We have H 6= G since s 6∈ H . Suppose H < H 0 . Choose x ∈ H 0 − H and

write x = sr k where 0 ≤ k < n. Then since r n−k ∈ H ,

x · r n−k = (sr k ) · r n−k = s · (r k r n−k ) = s · r n = s ∈ H 0

Hence r, s ∈ H 0 , so H 0 = G. Since H 0 was arbitrary, H is maximal.

(c) Let G = 〈x〉 be cyclic of order n. Then H ≤ G is maximal iff H = 〈x p 〉 for some
prime p dividing n.

Proof. Suppose H ≤ G is maximal. Write H = 〈x d 〉 where d ≥ 1 and d |n (by

Theorem 2.3.7). Since H 6= G, we have d > 1. Suppose d = ab with 1 < a, b < d .
Then 〈x d 〉 < 〈x a 〉 < G by the above note, contradicting the maximality of H .
Therefore d is prime.
Conversely, suppose H = 〈x p 〉 for some p|n. Then H 6= G since (p, n) = p > 1.
Suppose H < H 0 < G. Then H 0 = 〈x d 〉 for some d > 1 such that d |p but not
p|d —contradicting that p is prime. Therefore H is maximal.

E XERCISE 17. Let G be a nontrivial finitely generated group. Then G has a maximal
subgroup.

Proof. Let G be the set of all subgroups of G. Define a relation R from G to G by

H R g ⇐⇒ H < 〈H ∪ {g }〉 < G

Let F be a choice function for R. Note that for H ∈ G , H 6∈ dom(F ) iff either H = G or
H is maximal in G.
Let Ω denote the class of all ordinals and construct a recursion H : Ω → G as
follows: 

 1 if α = 0

〈H ∪ {F (H )}〉 if α = β+ and H ∈ dom(F )

β β β
Hα =


 Hβ if α = β and Hβ 6∈ dom(F )
+

if α is a limit ordinal

S
H
β<α β

It is easily verified by induction that α < β implies Hα ⊆ Hβ for all α, β ∈ Ω, and hence
by induction again that Hα ∈ G for all α ∈ Ω.

23
 Let γ be the Hartogs number of G . Since H |γ : γ → G cannot be injective, there
must exist ordinals δ < ε < γ such that Hδ = Hε . Note that δ < δ+ ≤ ε, hence

H δ ⊆ H δ+ ⊆ H ε ⊆ H δ

so Hδ = Hδ+ . By the definition of our recursion above, this implies that Hδ 6∈ dom(F ),
which means either Hδ = G or Hδ is maximal in G.
To rule out the first case, we prove (by induction) that Hα < G for all α ∈ Ω. Fix α
and suppose the result holds for all β < α. If α = 0, the result holds since H0 = 1
and 1 < G by hypothesis. If α = β+ , then both possible cases in our recursion yield
Hα < G. Finally, if α is a limit ordinal, suppose towards a contradiction that
[
G = Hα = Hβ
β<α

Since G is finitely generated, we have G = 〈g 0 , . . . , g n 〉 for some g 0 , . . . , g n ∈ G. Choose

β1 , . . . , βn < α such that g i ∈ Hβi for 1 ≤ i ≤ n. Since these subgroups form a chain,
we have H = Hβi = Hβ j for some 1 ≤ j ≤ n. Now g 0 , . . . , g n ∈ H , so G ≤ H —but this
S

contradicts the supposition that H < G. Thus Hα < G.

By induction then, our claim holds for all α ∈ Ω. In particular, Hδ is maximal
in G, so G has a maximal subgroup as desired.

Note that we present this direct proof instead of using Zorn’s Lemma because it well
illustrates both the principle behind Zorn’s Lemma as well as the natural way in
which the ordinals can be used to extend the counting process beyond the finite
(cf. the proof of this result for a finite group G).

E XERCISE 19.

(a) Q is divisible. Proof: For q ∈ Q and k 6= 0 arbitrary, set r = q/k, so r · k = q.

(b) No finite group is divisible.

Proof. Let G be an arbitrary nontrivial finite group. Write G = {g 0 , . . . , g n } and

set N = |g 0 | · · · |g n |. Then N > 0, and for all g ∈ G, g N = 1, so no nontrivial
element of G has an N -th root. It follows that G is not divisible.

Section 5

E XERCISE 3. Let H = 〈s, r 2 〉 in D 8 . Then H ∼

= V4 .

Proof. Note V4 = 〈 a, b | a 2 = 1 = b 2 , ab = ba 〉 and H = 〈 s, r 2 | s 2 = 1 = (r 2 )2 , sr 2 = r 2 s 〉.

Hence there is a natural isomorphism from V4 to H mapping a 7→ s and b 7→ r 2 .

24
E XERCISE 6. We compute the centralizers of each element in the following groups G:

(a) Let G = D 8 = 〈 s, r | s 2 = 1 = r 4 , sr = r −1 s 〉 = {1, r, r 2 , r 3 , s, r s, r 2 s, r 3 s}. Using the

group lattice and some computations, we have

C (1) = C (r 2 ) = D 8
C (r ) = C (r 3 ) = 〈r 〉
C (s) = C (r 2 s) = 〈s, r 2 〉
C (r s) = C (r 3 s) = 〈r s, r 2 〉

(c) Let G = S 3 = 〈(12), (123)〉 = {1, (12), (13), (23), (123), (132)}. We have

C (1) = S 3
C ((12)) = 〈(12)〉 C ((13)) = 〈(13)〉 C ((23)) = 〈(23)〉
C ((123)) = C ((132)) = 〈(123)〉

E XERCISE 8. We compute the normalizers of each element in the following groups G:

(a) Let G = S 3 as above. Then

N (1) = N (〈123〉) = N (S 3 ) = S 3
N (〈(12)〉) = 〈(12)〉 N (〈(13)〉) = 〈(13)〉 N (〈(23)〉) = 〈(23)〉

E XERCISE 10 (G ROUPS OF O RDER 4). If |G| = 4, then G ∼

= Z4 or G ∼
= V4 .

Proof. Suppose |G| = 4 but G ∼

6 Z4 . Then by Theorem 2.3.4(1), G is noncyclic, and
=
hence G contains no element of order 4. By Exercise 1.1.36 then, G ∼
= V4 .

E XERCISE 15. The group D 16 has three subgroups of order 8: H1 = 〈r 〉, H2 = 〈s, r 2 〉,

and H3 = 〈sr, r 2 〉. Trivially H1 ∼ = Z8 . By examining the subgroup lattice of D 16 , it
seems reasonable to guess that both H2 and H3 are isomorphic to D 8 .
Indeed, note that s 2 = 1 = (r 2 )4 , and s(r 2 ) = (r −1 )2 s = (r 2 )−1 s, and these relations
determine the structure of H2 , hence there is a natural isomorphism from D 8 to H2 .
Similarly, (sr )2 = 1 = (r 2 )4 and (sr )r 2 = (r −1 )2 (sr ) = (r 2 )−1 (sr ), so there is also a
natural isomorphism from D 8 to H3 .

Chapter 3
Section 1

25
E XERCISE 1. Let ϕ : G → H be a homomorphism and let E ≤ H . Then ϕ−1 (E ) ≤ G. If
E is normal, so is ϕ−1 (E ).

Proof. Write F = ϕ−1 (E ). Note 1G ∈ F since 1H ∈ E and ϕ(1G ) = 1H . Suppose x, y ∈ F

and write x 0 = ϕ(x) and y 0 = ϕ(y). By definition of F , {x 0 , y 0 } ⊆ E . Now

ϕ(x y −1 ) = ϕ(x)ϕ(y −1 )
= ϕ(x)ϕ(y)−1
= x 0 y 0−1 ∈ E

It follows that F ≤ G.
Suppose now that E E H . Let g ∈ G be arbitrary and x ∈ F . Then

ϕ(g xg −1 ) = ϕ(g )ϕ(x)ϕ(g )−1 ∈ E

since ϕ(x) ∈ E and E is normal. Thus g xg −1 ∈ F . Since g was arbitrary, F E G.

E XERCISE 3. Let A be an abelian group and B ≤ A. Then A/B is abelian.

Proof. The natural projection π : A → A/B is a surjective homomorphism.

Let G = D 6 and N = 〈r 〉. Then G is non-abelian and N E G. Since |G/N | = 2, G/N is

cyclic and hence abelian.

E XERCISE 4. Let G be a group and N E G. Then in G/N , (g N )α = g α N for all g ∈ G

and α ∈ Z.

Proof. Fix g ∈ G and α ∈ Z. Let π : G → G/N be the natural projection. Then

g α N = ϕ(g α ) = ϕ(g )α = (g N )α

E XERCISE 7. Define π : R2 → R by π(x, y) = x + y. Then π is a surjective homomor-

phism. The fibers of π are the lines y = c − x for c ∈ R. In particular, the kernel of π is
the line y = −x.

Proof. Trivial.

26
E XERCISE 14. Let G = Q/Z.

(a) Every coset in G contains exactly one representative q ∈ Q with 0 ≤ q < 1.

Proof. Let r + Z ∈ G where r ∈ Q, r ≥ 0. Let m ∈ Z be largest such that m ≤ r

and set q = r − m. Then 0 ≤ q < 1 and

r + Z = (q + m) + Z
= q + (m + Z)
= q +Z

Thus q is a desired representative for r + Z.

Suppose q + Z = q 0 + Z where q, q 0 ∈ Q and 0 ≤ q, q 0 < 1. Then q = q 0 + m for
some m ∈ Z, that is, m = q − q 0 . But we have 0 ≤ q − q 0 < 1, hence m = 0 and
q = q 0 . Thus the representative above is unique.

(b) All elements in G have finite order, but there exist elements of arbitrarily large
order.

Proof. Let q + Z ∈ G where q = m/n with m, n ∈ Z and n > 0. Then

n · (q + Z) = nq + Z = m + Z = Z

Hence |q + Z| ≤ n.
Note if (m, n) = 1, then |q + Z| = n. Thus for m = 1, |q + Z| → ∞ as n → ∞.

(c) G is the torsion subgroup of R/Z.

Proof. By (b), every element in Q/Z has finite order. Suppose α + Z ∈ R/Z and
|α + Z| = n. Then
n · (α + Z) = nα + Z = Z
In particular, nα = m for some m ∈ Z, so α = m/n ∈ Q. Thus α + Z ∈ Q/Z. This
establishes the claim.

(d) G is isomorphic to the multiplicative group of roots of unity in C× .

Proof. Define ϕ : G → C× by
m 2mπi
ϕ: + Z 7→ e n
n
It is verified that this a well defined, injective, surjective homomorphism.

27
E XERCISE 15. Let G be a divisible abelian group and N < G. Then G/N is divisible.

Proof. Since G is abelian, so is G/N , and since N 6= G, G/N 6= 1. Let g N ∈ G/N be

arbitrary and k ∈ Z, k 6= 0. Since G is divisible, there exists a ∈ G such that a k = g .
But then (aN )k = a k N = g N , hence g N has a k-th root in G/N .

E XERCISE 16. Let G = 〈S〉 and N E G. Then G = 〈 S 〉, where S = { s | s ∈ S }.

Proof. Let g ∈ G be arbitrary. Then since g ∈ 〈S〉, there exist s 1 , . . . , s n ∈ S such that
ε ε
g = s 11 · · · s nn (εi = ±1)

Therefore, since the natural projection from G to G is a homomorphism,

ε ε
g = s 11 · · · s nn
= s 1 ε1 · · · s n ε n ∈ 〈 S 〉

Since g was arbitrary, this shows G ≤ 〈 S 〉 as desired.

E XERCISE 22. Let G be a group and B be a nonempty set of normal subgroups of G.

Then B E G.
T

Proof. We know B ≤ G. Suppose x ∈ B and let g ∈ G be arbitrary. Then for

T T

any N ∈ B, x ∈ N , and hence g xg −1 ∈ N since N is normal. Since N was arbitrary,

g xg −1 ∈ B. Since g and x were arbitrary, B is normal.
T T

E XERCISE 23. Let G be a group and B be a nonempty set of normal subgroups of G.

Then 〈B〉 E G. (Here 〈B〉 denotes the join over B, that is, the smallest subgroup
of G which includes B.)
S

Proof. Suppose x ∈ 〈B〉 and let g ∈ G be arbitrary. Choose N ∈ B such that x ∈ N .

Then g xg −1 ∈ N since N is normal. But N ⊆ B ⊆ 〈B〉, hence g xg −1 ∈ 〈B〉. It
S

follows that 〈B〉 E G.

E XERCISE 24. Let G be a group, N E G and H ≤ G. Then N ∩ H E H .

Proof. We know N ∩ H ≤ H . If x ∈ N ∩ H and h ∈ H , then hxh −1 ∈ N since N is

normal in G, and hxh −1 ∈ H since H is a subgroup, hence hxh −1 ∈ N ∩ H .

28
 E XERCISE 25.
(a) Let G be a group and N ≤ G. Then N E G iff g N g −1 ⊆ N for all g ∈ G.

Proof. The forward direction is trivial. If g N g −1 ⊆ N for all g ∈ G, then for all
g ∈ G we have

N = (g g −1 )N (g g −1 ) = g (g −1 N g )g −1 ⊆ g N g −1

Hence g normalizes N for all g ∈ G, so N E G.

(b) Let G = GL 2 (Q), and define the subgroup and element

n µ1 n ¶ o µ
2
¶
N= |n∈Z g=
1 1

Note that for all n ∈ Z,

µ ¶ µ ¶µ ¶µ1 ¶
1 n −1 2 1 n 2
g g =
1 1 1 1
µ ¶
1 2n
= ∈N
1

Hence g N g −1 ⊆ N . But note that g does not normalize N , since not every
element in N has an even integer in entry (1, 2).
E XERCISE 26.
(a),(b) Recall that conjugation is an automorphism.
(c) Let G be a group, S ⊆ G, and N = 〈S〉. Then N E G iff g Sg −1 ⊆ N for all g ∈ G.

Proof. The forward direction is trivial. Suppose g Sg −1 ⊆ N for all g ∈ G. Let

x ∈ N be arbitrary. Then there exist s 1 , . . . , s n ∈ S such that
ε ε
x = s 11 · · · s nn (εi = ±1)

Since conjugation is a homomorphism, for any g ∈ G we have

g xg −1 = (g s 1 g −1 )ε1 · · · (g s n g −1 )εn ∈ N

Thus g N g −1 ⊆ N for all g ∈ G, so N E G.

(d) It follows immediately that if N = 〈x〉 is cyclic in G, then N is normal iff for all
g ∈ G, g xg −1 = x k for some k ∈ Z.
(e) Let G be a group, n ∈ Z, n > 0, and let N be the subgroup generated by all
elements in G of order n. Then N is normal.

Proof. Since conjugation is an automorphism, the conjugate of any element

of order n has order n. It follows from (c) that N is normal.

29
E XERCISE 29. Let G = 〈T 〉. Suppose S ⊆ G, N = 〈S〉, and N is finite. Then N E G iff
t St −1 ⊆ N for all t ∈ T .

Proof. The forward direction is trivial. Suppose t St −1 ⊆ N for all t ∈ T . Since N is

finite, NG (N ) = { g ∈ G | g Sg −1 ⊆ N } (Exercise 28). By hypothesis then, T ⊆ NG (N ).
Thus NG (N ) is a subgroup of G containing T , so G ⊆ NG (N ) and N E G.

E XERCISE 33. Write D 8 = 〈 r, s | r 4 = 1 = s 2 , r s = sr −1 〉. Using the lattice of subgroups

for D 8 and doing some computations (using Exercise 29), we find that the normal
subgroups of D 8 are 1, 〈r 2 〉, 〈r 〉, 〈s, r 2 〉, 〈r s, r 2 〉, and D 8 . The following isomorphisms
are easily verified:

D 8 /1 ∼= D8
2 ∼
D 8 /〈r 〉 = V4 (or D 4 )
D 8 /〈r 〉 ∼
= Z2 (or D 2 )
D 8 /〈s, r 2 〉 ∼
= Z2
2 ∼
D 8 /〈r s, r 〉 = Z2

E XERCISE 34. Write D 2n = 〈 r, s | r n = 1 = s 2 , r s = sr −1 〉 and let k be a positive integer

dividing n. Then 〈r k 〉 E D 2n and D 2n /〈r k 〉 ∼ = D 2k .

Proof. Write D 2k = 〈ρ, σ | ρ k = 1 = σ2 , ρσ = σρ −1 〉. Define

ϕ : D 2n → D 2k
s α r β 7→ σα ρ β (α, β ∈ Z)

Since ρ, σ satisfy in D 2k all of the relations satisfied by r, s in D 2n (note ρ n = (ρ k )n/k =

1n/k = 1), it follows that ϕ is a well defined map from D 2n to D 2k . It is easily seen that
ϕ is a surjective homomorphism.
If σα ρ β = 1, then 2|α and k|β, and conversely. Thus ker ϕ = 〈r k 〉. It follows that
〈r 〉 E G and G/〈r k 〉 ∼
k
= D 2k .

E XERCISE 36. If G/Z (G) is cyclic, then G is abelian.

Proof. Write Z = Z (G). Suppose G/Z is cyclic and fix g ∈ G such that G/Z = 〈g Z 〉.
Then for all x ∈ G, there exists α ∈ Z such that x ∈ (g Z )α = g α Z . In other words, for
all x ∈ G there exists α ∈ Z and z ∈ Z such that x = g α z.
Let x, y ∈ G be arbitrary, and write x = g α w and y = g β z where α, β ∈ Z and
w, z ∈ Z . Then we have

x y = (g α w)(g β z) = g α g β w z = g β g α zw = (g β z)(g α w) = y x

since w, z commute with all elements, and powers of g commute with each other.
Since x, y were arbitrary, it follows that G is abelian.

Another way to state this result is that G/Z (G) cannot be nontrivially cyclic (since,
when G is abelian, Z (G) = G so G/Z (G) ∼= 1).

30
E XERCISE 40. Let G be a group, N E G, and write G = G/N . For x, y ∈ G, x, y com-
mute in G iff x −1 y −1 x y ∈ N . (Note x −1 y −1 x y is called the commutator of x and y.)

Proof.

x y = y x ⇐⇒ x −1 y −1 x y = 1
⇐⇒ x −1 y −1 x y = 1
⇐⇒ x −1 y −1 x y ∈ N

E XERCISE 41. Let G be a group and set N = 〈 x −1 y −1 x y | x, y ∈ G 〉. Then N E G and

G/N is abelian. (Note N is called the commutator subgroup of G.)

Proof. For all x, y, g ∈ G,

g −1 (x −1 y −1 x y)g = (g −1 x −1 g )(g −1 y −1 g )(g −1 xg )(g −1 y g )

= (g −1 xg )−1 (g −1 y g )−1 (g −1 xg )(g −1 y g ) ∈ N

It follows from Exercise 26(c) that N E G. Since N contains all commutator elements
in G, it follows from Exercise 40 that G/N is abelian.

E XERCISE 43. Let G be a group and suppose A = { A i | i ∈ I } is a partition of G which

forms a group under the following operation: for A i , A j ∈ A , the product A i A j is the
element of A containing the product a i a j , for any a i ∈ A i and a j ∈ A j .
Then the identity A in A is a normal subgroup of G, and A ∼ = G/A.

Proof. Define π : G → A by mapping g → A i iff g ∈ A i . Then π is a well defined,

surjective homomorphism by the assumptions on A . Also ker π just consists of the
elements of A. Hence A E G and G/A ∼
=A.

This exercise shows that the quotient construction is in a certain sense ‘unique’. It is
the only construction on a partition of a group whose operation may be induced by
the operation of the group elements in the above manner.

Section 2

E XERCISE 2. We prove that the lattice of subgroups for S 3 in Section 2.5 is correct.

Proof. We first claim that the lattice is exhaustive. Indeed, if H is any subgroup of S 3 ,
then by Lagrange’s Theorem we know the order of H is one of 1, 2, 3, or 6. Cases
|H | = 1 and |H | = 6 are trivial. If |H | = 2, then H is cyclic and generated by one of the
three elements of order 2, that is, H is one of

〈(12)〉 〈(13)〉 〈(23)〉

Similarly if |H | = 3, then H must be cyclic and equal to 〈(123)〉. This shows that the
lattice is exhaustive.

31
 Clearly any two of the distinct subgroups of order 2 have trivial intersection. Also,
by Lagrange’s Theorem, each of these subgroups have trivial intersection with the
subgroup of order 3 since (2, 3) = 1. This shows that the intersections on the lattice
are correct.
Finally, it is easily verified that any two distinct nontrivial subgroups join to all
of S 3 . Hence the joins on the lattice are correct.

E XERCISE 4. Suppose |G| = pq, with p, q prime. Then Z (G) = 1 or G is abelian.

Proof. Let H = G/Z (G) and suppose Z (G) > 1. If Z (G) = G, then trivially G is abelian.
If Z (G) 6= G, then by Lagrange’s Theorem, |Z (G)| is either p or q, and thus the same
is true of |H |. But then H is cyclic, so again G is abelian by Exercise 3.1.36.

E XERCISE 6. Suppose H ≤ G, g ∈ G, and the right coset H g is equal to some left coset
of H . Then g H = H g .

Proof. Suppose H g = g 0 H for some g 0 ∈ G. Then g H and g 0 H both contain g , hence

g H = g 0H = H g .

E XERCISE 8. Suppose H and K are finite subgroups of G whose orders are relatively
prime. Then H ∩ K = 1

Proof. By Lagrange’s Theorem since |H ∩ K | must divide both |H | and |K |.

E XERCISE 9 (C AUCHY ’ S T HEOREM ). Let G be a finite group and suppose p is a prime

dividing |G|. Then there exists an element in G of order p.

Proof. First define

P = { (x 1 , . . . , x p ) | x i ∈ G ∧ x 1 · · · x p = 1 }

(a) Note that |P | = |G|p−1 since, when constructing a tuple (x 1 , . . . , x p ), we have

|G| possible choices for each x i where 1 ≤ i ≤ p − 1, but once those elements
are chosen, we must choose x p = (x 1 · · · x p−1 )−1 . Thus p divides |P |.
Define a relation ∼ on P by letting α ∼ β iff α is a cyclic permutation of β.
(b) Note P is closed under cyclic permutations. Indeed, if (x 1 , . . . , x p ) ∈ P , then
(x 1 · · · x p−1 )x p = 1, so x p (x 1 · · · x p−1 ) = 1, so (x p , x 1 , . . . , x p−1 ) ∈ P . The general
case now follows by induction.
(c) Note ∼ is reflexive since the identity permutation is cyclic; it is symmetric
since the inverse of a cyclic permutation is also cyclic; and it is transitive since
the composite of two cyclic permutations is also cyclic. Therefore ∼ forms an
equivalence relation.
(d) We claim that an equivalence class contains a single element iff that element
is of the form (x, . . . , x), with x p = 1. Indeed, if x p = 1, then (x, . . . , x) is in P ,
and it is preserved under any cyclic permutation. Conversely, if (x 1 , . . . , x p ) is
in P and is preserved under any cyclic permutation, then by applying powers
of the permutation (of indices) (1 · · · p), it follows that x 1 = · · · = x p . So the
tuple is of the desired form with x = x 1 , and x p = 1.

32
 (e) We claim that each equivalence class has order 1 or p. To see this, let E be
an arbitrary equivalence class and note that the set of cyclic permutations of
elements of E forms a group H of order p under composition. Moreover, this
group acts on E . Now fix a tuple t ∈ E and consider the stabilizer S of t in H .
By Lagrange’s Theorem, since p is prime, either S = 1 or S = H . If S = H , then
E has order 1. If S = 1, then distinct cyclic permutations give rise to distinct
images of t . So H is in bijective correspondence with E , and E has order p.
It follows now that |G|p−1 = k +pd , where k is the number of 1-element classes
and d is the number of p-element classes.
(f) Since p divides |G|p−1 , p divides k + pd , and so p divides k. We know k > 0
since {(1, . . . , 1)} is a 1-element equivalence class. But this means k > 1, so there
must exist a nonidentity element x ∈ G with x p = 1, and |x| = p.

Note that this theorem is a generalization of Exercise 1.1.31, and the proof is in some
ways a generalization of the proof used in that exercise.

E XERCISE 11. Suppose H ≤ K ≤ G. Then |G : H | = |G : K ||K : H |.

Proof. If |K : H | = ∞, then there are infinitely many cosets of H in K , and hence in G,

so |G : H | = ∞ and equality holds.
If |G : K | = ∞, then there are infinitely cosets of K in G. Note that for all g ∈ G,
g H ⊆ g K , that is, each coset of K contains a G-coset of H . Since the cosets of K
are pairwise disjoint, these G-cosets of H are pairwise distinct (and disjoint). Hence
|G : H | = ∞ and equality holds in this case as well.
Suppose now that |G : K | and |K : H | are both finite. We claim that the number
of G-cosets of H contained in each coset of K is equal to |K : H |. Indeed, let g ∈ G be
arbitrary and define the map

π : k H 7→ g k H (k ∈ K )

Note for all k ∈ K , k H ⊆ K , so g k H ⊆ g K . Thus π maps cosets of H in K to cosets

of H contained in g K . This map is injective, since if g k H = g k 0 H , then k H = k 0 H .
Also, if g 0 H ⊆ g K , then in particular g 0 ∈ g K , so g 0 = g k ∗ for some k ∗ ∈ K and g 0 H =
g k ∗ H = π(k ∗ H ). Thus π witnesses a bijection between the set of K -cosets of H and
the set of G-cosets of H contained in g K . Since g was arbitrary, this proves our claim.
It now follows that |G : H | is finite, and indeed the above equality holds.

Note if we write this equality as

|G : H |
|G : K | =
|K : H |

this suggests that the quotient G/K might be isomorphic to the quotient (G/H )/(K /H ).
Indeed, this is so (see the Third Isomorphism Theorem).

33
E XERCISE 15. Let G = S n , fix i ∈ {1, . . . , n}, and let G i be the stabilizer of i . Then
Gi ∼
= S n−1 .

Proof. Define σ = (i n). Then

Gi ∼
= σG i σ = G n ∼
−1
= S n−1

where the first isomorphism is obtained by conjugation and the second is obtained
by restriction.

E XERCISE 21. Neither Q nor Q/Z have proper subgroups of finite index.

Proof. Recall that if G is divisible, then G cannot be finite (Exercise 2.4.19(b)). Also
if G is divisible and H is a proper normal subgroup of G, then G/H is also divisible
(Exercise 3.1.15). Thus a divisible group cannot have proper normal subgroups of
finite index.
The result now follows from the fact that Q is abelian and divisible (Exercise
2.4.19(a)).

E XERCISE 22 (E ULER ’ S T HEOREM ). Fix n and let a ∈ Z be relatively prime to n. Then

a ϕ(n) ≡ 1 mod n.

Proof. Consider G = (Z/nZ)× . Recall a ∈ G since (a, n) = 1, and |G| = ϕ(n). Hence by
Lagrange’s Theorem, |a| divides ϕ(n), so a ϕ(n) = 1—that is, a ϕ(n) ≡ 1 mod n.
100
E XERCISE 23. We compute the last two digits of 33 using Euler’s Theorem. Note
100
that this is equivalent to computing 33 mod 100. By Euler’s Theorem, note that
100
3ϕ(100) mod 100 = 1, hence we need only compute 33 mod ϕ(100) mod 100.
100
Write r = 3 mod ϕ(100). Note ϕ(100) = ϕ(2 · 52 ) = 2(1) · 5(4) = 40. Thus to
2

compute r , we may use Euler’s Theorem again to note r = 3100 mod ϕ(40) mod 40.
Now ϕ(40) = ϕ(23 · 5) = 22 (1) · 1(4) = 16, and 100 mod 16 = 4. So r = 34 mod 40 = 1.
Thus our answer is 31 mod 100 = 3–that is, the last two digits are 03.

Section 3

E XERCISE 2 (L ATTICE T HEOREM ). Let G be a group and N E G. For any A ≤ G, let

A = A/N = { aN | a ∈ A }

Then the map A 7→ A forms a bijective correspondence between the set of subgroups
of G containing N and the set of subgroups of G = G/N . Moreover, this bijection
preserves lattice structure: for A, B ≤ G with N ≤ A ∩ B ,

(1) A ≤ B iff A ≤ B
(2) if A ≤ B then |B : A| = |B : A|
(3) 〈A, B 〉 = 〈 A, B 〉
(4) A ∩ B = A ∩ B
(5) A E G iff A E G

34
Proof. Let π : G → G be the natural projection. Then for N ≤ A ≤ G,

A = π[A] ≤ π[G] = G

Hence A ≤ G, so the map is defined on the desired sets.

Fix A, B ≤ G with N ≤ A ∩B . Note that if aN = bN for any a ∈ A and b ∈ B , then in
particular a ∈ B (since N ≤ B ) and b ∈ A (since N ≤ A). Hence if A = B , then A = B .
In other words, the mapping above is injective.
Finally, the mapping is surjective. If A ∗ ≤ G is arbitrary, set A = π−1 [A ∗ ]. Then
A ≤ G, and N ≤ A since N ∈ A ∗ . Also A = A ∗ by definition of π−1 .
Properties (1) and (3)–(5) are immediate by direct computation with elements.
To prove (2), suppose N ≤ A ≤ B and define a map from the coset space B /A to the
coset space (B /N )/(A/N ) as follows:

f : B /A → (B /N )/(A/N )
b A 7→ bN (A/N )

By definition
bN (A/N ) = { (bN )(aN ) | a ∈ A } = { baN | a ∈ A }
Note f is well-defined. Indeed, if b A = b 0 A, then b = b 0 a 0 for some a 0 ∈ A, hence
baN = b 0 (a 0 a)N for all a ∈ A, and so bN (A/N ) ⊆ b 0 N (A/N ). The converse follows by
symmetry.
Also f is injective. If bN (A/N ) = b 0 N (A/N ), then bN = b 0 aN ⊆ b 0 A for some
a ∈ A. In particular b ∈ b A ∩ b 0 A, hence b A = b 0 A.
Clearly f is surjective. Hence f is a bijection and |B : A| = |B : A|.

E XERCISE 3. Suppose H E G and |G : H | = p for some prime p. Then for all K ≤ G,

either K ≤ H or else G = H K and |K : K ∩ H | = p.

Proof. Let K ≤ G be arbitrary. Note H ≤ H K ≤ G, hence

p = |G : H | = |G : H K ||H K : H |

Suppose K 6≤ H . Then |H K : H | > 1, hence |H K : H | = p and |G : H K | = 1 (since p is

prime), and thus G = H K .
Now by the Diamond Isomorphism Theorem, G/H ∼ = K /(K ∩ H ), so in particular
|K : K ∩ H | = |G : H | = p.

E XERCISE 4. Let A and B be groups, C E A and D E B . Then C × D E A × B and

(A × B )/(C × D) ∼
= (A/C ) × (B /D)

Proof. Define the mapping

ϕ : (A × B ) → (A/C ) × (B /D)
(a, b) 7→ (aC , bD)

Then it is immediate that ϕ is a surjective homomorphism with ker ϕ = C × D. The

desired result now follows from the First Isomorphism Theorem.

35
E XERCISE 7. Suppose G = M N where M E G and N E G. Then

G/(M ∩ N ) ∼
= (G/M ) × (G/N )
Proof. Define the map

ϕ : G → (G/M ) × (G/N )
g 7→ (g M , g N )

It is immediate that ϕ is a homomorphism with ker ϕ = M ∩ N .

We claim that ϕ is surjective. Indeed, since G = M N = N M , we have G/M = N /M
(coset space) and G/N = M /N (coset space). Thus values

ϕ(mn) = ϕ(m)ϕ(n) = (M , mN )(nM , N ) = (nM , mN )

exhaust (G/M ) × (G/N ) as m and n vary.

The desired result now follows from the First Isomorphism Theorem.

Note that this result can be seen visually by looking at the (partial) subgroup lattice.
The ‘diamond’ formed by G/(M ∩ N ) is reobtained by taking the cross product of the
‘lines’ G/M and G/N .

E XERCISE 9 (R ESTRICTING S YLOW p -S UBGROUPS ). Let G be a group with |G| = p α m

where p does not divide m. Suppose P is a Sylow p-subgroup of G (that is, |P | = p α )
and N E G with |N | = p β n where p does not divide n. Then P ∩ N is a Sylow p-
subgroup of N (that is, |P ∩ N | = p β ) and |P N /N | = p α−β .

Proof. Note that if |P ∩ N | = p β , then by the Diamond Isomorphism Theorem and

Lagrange’s Theorem, we have
|P | pα
|P N /N | = |P /(P ∩ N )| = = β = p α−β
|P ∩ N | p

Hence it suffices to prove |P ∩ N | = p β .

Let κ = |P ∩ N |. By Lagrange’s Theorem, κ|p α and κ|p β n, hence κ|p β . Now let
λ = |P N |. Recall
|P ||N | p α p βn
λ= =
|P ∩ N | κ
But λ|p α m by Lagrange’s Theorem. Since p cannot divide m, we must have p β |κ.
Hence κ = p β as desired.

Section 4

E XERCISE 1. Let G be an abelian simple group. Then G ∼

= Z p for some prime p.

Proof. By definition |G| > 1. Fix x ∈ G, x 6= 1 and set N = 〈x〉. Since G is abelian,
any subgroup is normal and so 1 < N E G. But this implies N = G since G is simple,
so G is cyclic. By Theorem 2.3.7, G cannot have infinite or finite composite order
because it does not have any nontrivial proper subgroups. Therefore G ∼ = Z p for
some prime p.

36
E XERCISE 4. Let G be a finite abelian group and suppose n > 0 divides the order of G.
Then G has a subgroup of order n.

Proof. If n is prime then the result follows from Cauchy’s Theorem.

If n is composite, write n = pm for a prime p and m > 0. By Cauchy’s Theorem
choose a subgroup N with |N | = p. Note that N E G since G is abelian. Set G =
G/N . By Lagrange’s Theorem, |G| = |G|/p, hence |G| < |G|, and since n divides |G|,
m must divide |G|. By induction choose a subgroup H ≤ G with |H | = m. Let H =
π−1 [H ] where π : G → G denotes the natural projection. By the Lattice Isomorphism
Theorem and Lagrange’s Theorem, N ≤ H ≤ G and

|H | = |N ||H : N | = |N ||H | = pm = n

Therefore H is a subgroup of G of order n, as desired.

E XERCISE 5. Let G be a solvable group. If H ≤ G, then H is solvable. If N E G, then

G/N is solvable.

Proof. Let 1 = G 0 E · · · E G n = G be a chain of subgroups in G such that G i +1 /G i is

abelian for 0 ≤ i < n. Suppose H ≤ G. Define Hi = G i ∩ H for 0 ≤ i ≤ n. Then trivially

1 = H0 E · · · E Hn = H

Let 0 ≤ i < n. We claim Hi +1 /Hi is abelian. Indeed, let A = Hi +1 and B = G i . Then

A ≤ NG (B ) since G i E G i +1 , A ∩ B = Hi , and AB ≤ G i +1 . Therefore by the Diamond
Isomorphism Theorem,

Hi +1 /Hi = A/(A ∩ B ) ∼
= AB /B ≤ G i +1 /G i

Therefore Hi +1 /Hi is abelian. Since i was arbitrary, H is solvable.

Now suppose N E G. For H ≤ G, set H = π[H ], where π : G → G/N is the natural
projection. Then by the Lattice Isomorphism Theorem, we have

1 = G0 E · · · E Gn = G

By the proof of the Third Isomorphism Theorem, for all 0 ≤ i < n, G i +1 /G i ∼

= G i +1 /G i
and hence is abelian.

E XERCISE 6,7 ( J ORDAN -H ÖLDER T HEOREM PART I). Let G be a nontrivial finite group.
Then G has a composition series.

Proof. If G has no nontrivial proper normal subgroups, then G is simple and 1 E G

is a composition series for G.
If not, choose 1 < N / G. Since |N | < |G|, by induction we may assume N has a
composition series 1 = N0 E · · · E N j = N . Set G = G/N . Then |G| < |G|, so again by
induction we may assume G has a composition series 1 = G 0 E · · · E G k = G.
Set G i = π−1 [G i ] for 0 ≤ i ≤ k, where π : G → G is the natural projection. By the
Lattice Isomorphism Theorem, we have N = G 0 E · · · E G k = G, and by the Third

37
Isomorphism Theorem, G i +1 /G i ∼
= G i +1 /G i , and hence is simple, for 0 ≤ i < k. Thus
we have a composition series for G given by

1 = N0 E · · · E N j = N = G 0 E · · · E G k = G

Note that this proof also shows that if H E G, there exists a composition series for G
containing H as a term. Indeed, if H = 1 or H = G, this is trivial. Otherwise, combine
a composition series for H with the preimage of a composition series for G/H under
the natural projection to obtain a composition series for G which includes H .
An analogous note also holds for solvable groups.

Section 5

N OTE .Every permutation in S n can be written as a product of transpositions. To

prove this, first claim that this holds for any m-cycle:

(a 1 · · · a m ) = (a 1 a m ) · · · (a 1 a 2 )

Indeed, this follows by induction. For m = 1 the result is trivial. If the result holds for
fixed m ≥ 1, note that

(a 1 · · · a m+1 ) = (a 1 a m+1 )(a 1 · · · a m ) by direct verification

= (a 1 a m+1 )[(a 1 a m ) · · · (a 1 a 2 )] by induction
= (a 1 a m+1 ) · · · (a 1 a 2 )

Thus the result holds for m + 1. By induction the result holds for all m.
Now for any σ ∈ S n , σ may be written as a product of cycles (cycle decomposition
algorithm). Since each of these cycles may be written as a product of transpositions
by the above, the result holds for σ.

E XERCISE 3. S n = 〈 (i i + 1) | 1 ≤ i < n 〉

Proof. Let S denote the set on the right. Claim that for any 1 ≤ i < j ≤ n, (i j ) ∈ S.
Indeed, this follows by induction on j . For j = 1 the result is trivial. If the result
holds for j < n, note that (i j + 1) = (i j )( j j + 1)(i j ) ∈ S.
Since any element of S n is a product of transpositions, the result follows.

E XERCISE 4. S n = 〈(12), (1 · · · n)〉 for n ≥ 2.

Proof. Let S denote the set on the right. By assumption (12) ∈ S. Suppose 1 < i < n.
Set ρ = (1 · · · n)i −1 . Then (i i + 1) = ρ(12)ρ −1 ∈ S.
The result now follows from Exercise 3.

38
E XERCISE 7. The group of rigid motions (rotations) of a regular tetrahedron in 3-
space is isomorphic to A 4 .

Proof. Let G be the group of rigid motions. By letting G act on numbered faces of the
tetrahedron, we obtain a permutation representation ϕ : G → S 4 . Note |ϕ[G]| ≤ 12
since any rigid motion must take face 1 to one of four possible faces, and at any one
of these must put it in one of three possible positions. And |ϕ[G]| ≥ 12 since each
one of these configurations is distinct and is realized by a rigid motion in G. Hence
we must have |ϕ[G]| = 12.
We claim that ϕ[G] ⊆ A 4 . First note that G is generated by two rotations ρ 1 and ρ 2
through 2π/3 radians about axes passing through any two distinct vertices and the
centers of their respective opposing faces. Therefore since ϕ is a homomorphism, it
suffices to observe that ϕ(ρ 1 ) and ϕ(ρ 2 ) can be written as even permutations.
Since ϕ[G] ⊆ A 4 and |ϕ[G]| = 12 = |A 4 |, ϕ[G] = A 4 and so ϕ : G ∼
= A4.

N OTE .Recall that groups of rigid motions of 2-dimensional regular n-gons in 3-space
are not in general isomorphic to subgroups of A n (e.g. D 6 ∼
= S 3 ). Intuitively it makes
sense that the groups of rigid motions of a 3-dimensional object in 3-space (acting
on a set of n elements) might be, since it does not contain reflections.

E XERCISE 8,9,14. The lattice of subgroups for A 4 in Figure 8 is correct.

Proof. We first prove exhaustiveness. Note |A 4 | = 12 = 2 · 2 · 3, hence by Lagrange’s

Theorem the only possible subgroup orders are 1, 2, 3, 4, 6, 12. Orders 1 and 12 are
trivial. There are no subgroups of order 6 by Exercise 7 and the remarks on p. 92.
If a subgroup has order 2, it must be cyclic and generated by one of the even
permutations in S 4 of order 2: (12)(34), (13)(24), or (14)(23). Similarly if a subgroup
has order 3, then it must be cyclic and generated by one of these (even) 3-cycles in S 4 :
(123), (124), (134), or (234). If a subgroup has order 4, it cannot contain an element of
order 4 (since a 4-cycle is odd). Hence it must contain, and be generated by, at least
two distinct even permutations of order 2. However, it is easily checked that any
of the two distinct even permutations above generate the same subgroup. Hence
there is just one subgroup of order 4, namely 〈(12)(34), (13)(24)〉. This establishes
exhaustiveness of the lattice.
We now establish (nontrivial) intersections. Clearly any two distinct subgroups
of order 2 have trivial intersection, and similarly for the subgroups of order 3. By
Lagrange’s Theorem, the intersection of a subgroup of order 2 or 4 with a subgroup
of order 3 is trivial since (2, 3) = (4, 3) = 1. Finally, the intersection of the subgroup of
order 4 with any subgroup of order 2 follows from the inclusions noted above.
The joins are established similarly. Hence the lattice is correct.

Note that the subgroup of order 4 is isomorphic to V4 since it has no element of

order 4. Also it must be normal since it is the only subgroup of order 4. It is thus
immediate that
1 E 〈(12)(34)〉 E 〈(12)(34), (13)(24)〉 E A 4

39
By the lattice, the factors in this chain are cyclic of orders 2 or 3, and hence are
abelian simple. It follows that this is a composition series for A 4 and moreover that
A 4 is solvable.

Chapter 4
Section 1

N OTE .If G acts on A and a ∈ A, then by Lagrange’s Theorem and Proposition 2,

|G| = |G : G a ||G a | = |O a ||G a |

where O a is the orbit of a under G, and G a is the stabilizer of a in G. This result is

sometimes called the Orbit-Stabilizer Theorem.

E XERCISE 1. Let G act on A and suppose b = g · a for some a, b ∈ A and g ∈ G. Then

G b = gG a g −1 . If G is transitive on A, then the kernel of the action is g ∈G gG a g −1
T

Proof. If x ∈ G a , then for g ∈ G,

(g xg −1 )(b) = (g x)(g −1 (b)) = (g x)(a) = g (xa) = g (a) = b

hence g xg −1 ∈ G b . Thus gG a g −1 ⊆ G b .
Conversely, if x ∈ G b , then by the above (with g −1 ), g −1 xg ∈ G a , hence

x = g (g −1 xg )g −1 ∈ gG a g −1

Therefore G b = gG a g −1 as desired. The second claim is then immediate since the

kernel of the action is just the intersection of the stabilizers.

E XERCISE 3. Let G be an abelian, transitive subgroup of S A . Then for all σ ∈ G − 1,

σ(a) 6= a for all a ∈ A.

Proof. If a ∈ A, then gG a g −1 = G a for all g ∈ G since G is abelian. By Exercise 1 then,

g (a) = a implies that g is in the kernel of the action, that is, g = 1.

E XERCISE 7. Let G be a transitive permutation group on A. Call B ⊆ A a block iff B is

nonempty and for all σ ∈ G, σ[B ] = B or σ[B ] ∩ B = ; (where σ[B ] = { σ(b) | b ∈ B }).
Call G primitive on A iff the only blocks in A are the trivial ones (the singletons, and
the entire set).

(a) If B is a block and b ∈ B , then G b ≤ G B ≤ G, where G B = { σ ∈ G | σ[B ] = B }.

Proof. Clearly 1 ∈ G B , and if σ, τ ∈ G B , then τ−1 [B ] = B , hence (στ−1 )[B ] =

σ[τ−1 [B ]] = σ[B ] = B . Thus G B ≤ G. If σ ∈ G b , then b = σ(b) ∈ B ∩ σ[B ], hence
σ[B ] = B (since B is a block) and σ ∈ G B . Thus G b ≤ G B .

(b) If B is a block, then the set B = { σ[B ] | σ ∈ G } partitions A.

40
 Proof. Each set in B is nonempty since B is nonempty. Fix b ∈ B . Since G is
transitive, for any a ∈ A there exists σ ∈ G such that a = σ(b) ∈ σ[B ]. Hence
A = B. We claim that σ[B ] 6= τ[B ] implies σ[B ] ∩ τ[B ] = ;. Indeed, suppose
S

a ∈ σ[B ] ∩ τ[B ], so a = σ(b) = τ(b 0 ) for some b, b 0 ∈ B . Then

b 0 = τ−1 (a) = τ−1 (σ(b)) = (τ−1 σ)(b) ∈ (τ−1 σ)[B ]

Thus b 0 ∈ B ∩ (τ−1 σ)[B ], so B = (τ−1 σ)[B ] and τ[B ] = σ[B ].

(c) S 4 is primitive on A = {1, 2, 3, 4}, but D 8 is not primitive on the vertices of the
square.

Proof. Suppose X ⊆ A is nontrivial (in the above sense). Choose x ∈ X and

y ∈ A − X . Let σ = (x y) ∈ S 4 . Then σ[X ] 6= X but σ[X ] ∩ X is nonempty. Thus
X is not a block. Since X was arbitrary, S 4 is not primitive.
On the contrary, if we let A label the vertices of the square in clockwise order,
then the pair X = {1, 3} of diagonally opposing vertices is a block under D 8 .

(d) G is primitive iff G a is maximal for all a ∈ A.

Proof. If G is not primitive, there exists a nontrivial block B in A. Fix b, b 0 ∈ B

with b 6= b 0 and a ∈ A − B . We know G b ≤ G B ≤ G. Since G is transitive, there
exists σ ∈ G with σ(b) = b 0 . Note σ ∈ G B − G b , hence G b < G B . Similarly there
exists τ ∈ G with τ(b) = a, so τ 6∈ G B and G B < G. It follows that G b is not
maximal.
If G a is not maximal, then there exists G a < H < G. Let B be the orbit of a
under H . Since G a < H , B contains elements other than a. Fix σ ∈ G − H .
If σ(a) = τ(a) for some τ ∈ H , then τ−1 σ ∈ G a ⊆ H , so σ ∈ H , contradicting
the choice of σ. Hence σ(a) ∈ A − B and B 6= A. We claim that B is a block.
Indeed, suppose σ[B ] ∩ B is nonempty, so there exist b, b 0 ∈ B with b 0 = σ(b).
Choose τ, τ0 ∈ H with b = τ(a) and b 0 = τ0 (a). Then τ0 (a) = σ(τ(a)) = (στ)(a),
so τ0−1 (στ) = τ0−1 στ ∈ G a ⊆ H . But this implies that σ ∈ H and σ[B ] = B . It
follows that B is a block as claimed. Since B is nontrivial, it follows that G is
not primitive.

Informally, a block may be thought of as a set of related elements which always

‘move together’ as a unit under the action of a group. Hence the elements of a block
may be ‘identified’ to induce a natural action of the group on the blocks themselves.
In general, stabilizers under the induced action will be larger than corresponding
stabilizers under the original action because the stabilizer of a block includes the
stabilizers of the individual elements in the block as well as any permutations which
permute the elements of the block amongst each other. Also, by the Orbit-Stabilizer
Theorem above, this means orbits generally get smaller.

41
E XERCISE 9 (O RBIT-S TABILIZER ). Let G act transitively on the finite set A and H E G.
Let O 1 , . . . ,O r denote the distinct orbits of H on A.
(a) G acts transitively on the orbits where g ·O i = { g · a | a ∈ O i }, and all the orbits
have the same cardinality.

Proof. Write O i = { h · a | h ∈ H }. Then

g · O i = { g · (h · a) | h ∈ H }
= {gh ·a | h ∈ H }
= { hg · a | h ∈ H } since g H = H g
= { h · (g · a) | h ∈ H } = O j
where O j is the orbit of g · a under H . It follows that G acts transitively on the
orbits since G acts transitively on A.
Since g is a permutation of A, all the orbits have the same cardinality.

(b) Suppose a ∈ O 1 . Then |O 1 | = |H : H ∩G a | and r = |G : HG a |.

Proof. The first claim follows from the Orbit-Stabilizer Theorem (Proposition 2)
since H a = H ∩ G a . The second claim follows similarly by noting that HG a is
precisely the stabilizer of O 1 in the action of G on the orbits.

Section 2

E XERCISE 8. Suppose H ≤ G and |G : H | = n. Then there exists K ≤ H with K E G

and |G : K | ≤ n!.

Proof. Let G act on the coset space A = G/H , let π : G → S A be the permutation
representation and set K = ker π. Then K ≤ H , K E G, and by the First Isomorphism
Theorem,
|G : K | = |G/K | = |ϕ[G]| ≤ |S A | = n!

E XERCISE 10 (G ROUPS OF O RDER 6). If |G| = 6, then G ∼

= Z6 or G ∼
= S 3 (∼
= D 6 ).

Proof. Since |G| = 3 · 2, by Cauchy’s Theorem there exist elements r, s ∈ G with |r | = 3

and |s| = 2. Set H = 〈r s〉. If G is abelian, then H contains the distinct elements 1,
r = (r s)4 , s = (r s)3 , and r s. By Lagrange’s Theorem then, |H | = 6, so H = G, which
means G is cyclic and G ∼ = Z6 .
If G is not abelian, set K = 〈s〉 and let G act on the coset space G/K . Note that this
gives a permutation representation into S 3 since by Lagrange’s Theorem |G/K | = 3.
We claim that this representation is faithful. Indeed, note G/K is given by
〈s〉 = {1, s} r 〈s〉 = {r, r s} r 2 〈s〉 = {r 2 , r 2 s}
If g is in the kernel of the action, then in particular g must stabilize 〈s〉, so g = 1 or
g = s. But s does not stabilize r 〈s〉 since sr 6= r (s 6= 1) and sr 6= r s (G is not cyclic).
So we must have g = 1. Therefore the kernel of the action is trivial, and G ∼ = S3.

42
E XERCISE 11. Let G be a finite group and πG : G → SG denote the left regular action
of G. If x ∈ G with |x| = n and |G| = mn, then π(x) is a product of m n-cycles. Hence
π(x) is an odd permutation iff |x| is even and |G|/|x| is odd.

Proof. Set X = 〈x〉. Then |G : X | = m, hence there exist elements g 1 , · · · , g m ∈ G with

g 1 = 1 such that the n-element right cosets X , X g 2 , · · · , X g m partition G. But these
are just the orbits under the left action of X on G. In particular, we may write

π(x) = (1 x · · · x n−1 )(g 2 xg 2 · · · x n−1 g 2 ) · · · (g m xg m · · · x n−1 g m )

which is a product of m n-cycles as claimed.

Recall an n-cycle is even iff n is odd. Hence π(x) is even iff n is odd or else m is
even, or equivalently π(x) is odd iff n is even and m is odd. This completes the
proof.

E XERCISE 12. Let G be a finite group and π : G → SG be the left regular action. If
π[G] contains an odd permutation, then G has a subgroup of index 2.

Proof. Recall |SG : AG | = 2. Since π[G] contains an odd permutation, π[G] 6≤ AG .

Hence SG = π[G]AG and by the Diamond Isomorphism Theorem,

|π[G] : π[G] ∩ AG | = |SG : AG | = 2

(See Exercise 3.3.3.) Therefore π−1 [π[G] ∩ AG ] has index 2 in G.

E XERCISE 13. Let G be a group with |G| = 2k where k is odd. Then G has a subgroup
of index 2.

Proof. Since G has even order, there exists x ∈ G of order 2. But then by Exercise 11,
πG (x) is an odd permutation (where πG denotes the left regular representation of G).
Hence G has a subgroup of index 2 by Exercise 12.

Note that the previous three exercises illustrate a recurring use of group actions:
pulling information about elements and subgroups of S n back into arbitrary groups.
In this case we have used information about the existence of A n in S n to obtain
information about more general groups.

E XERCISE 14. Let |G| = n with n composite. Suppose for each positive k|n there
exists a subgroup of G of order k. Then G is not simple.

Proof. Note n > 1. Let p be the smallest prime dividing n and write k = n/p. Then
1 < k < n. By assumption there exists a subgroup K with |K | = k, and 1 < K < G.
By Lagrange’s Theorem, |G : K | = p, so K E G by Corollary 5. It follows that G is not
simple.

Section 3

43
E XERCISE 2. We find conjugacy class decompositions.

(a) Let G = D 8 = 〈 r, s | s 2 = 1 = r 4 , sr = r −1 s 〉. We know that Z (D 8 ) = 〈r 2 〉, hence

two conjugacy classes are {1} and {r 2 }. Now 〈r 〉 ≤ C (r ), but C (r ) 6= D 8 since
s 6∈ C (r ) (sr s = r 3 ), so C (r ) = 〈r 〉. Since |〈r 〉| = 4, by Lagrange’s Theorem we
have |G : 〈r 〉| = 2, so by the Orbit-Stabilizer Theorem the conjugacy class of r
consists of the two elements {r, r 3 }. It is similarly verified that C (s) = 〈s, r 2 〉, so
the conjugacy class of s contains the two elements {s, sr 2 }, and C (sr ) = 〈r s, s 2 〉,
so the conjugacy class of sr contains the two elements {sr, sr 3 }. This gives a
complete partition. The class equation for D 8 is 8 = 1 + 1 + 2 + 2 + 2.
(b) Let G = A 4 . Trivially Z (A 4 ) = 1. Using the lattice and the fact that groups of
order 4 are abelian, it is easily checked that C ((12)(34)) = 〈(12)(34), (13)(24)〉.
Thus, as above, the conjugacy class of (12)(34) consists of 3 elements. Since
conjugates must have the same cycle shape, the only possible conjugates are
(12)(34), (13)(24), and (14)(23). Hence these are exhaustive.
Similarly C ((123)) = 〈(123)〉, hence the conjugacy class containing (123) has
4 elements. These must be 3-cycles. It is checked by direct computation that
these are (123), (142), (134), and (243). This leaves the elements (132), (124),
(143), and (234) in the remaining conjugacy class.
Thus the class equation for A 4 is 12 = 1 + 3 + 4 + 4.

N OTE .For the following exercise we first observe the following result: If A and B are
groups, then the conjugacy classes in A × B are precisely the direct products of the
conjugacy classes of A with the conjugacy classes of B .

Proof. Trivial since multiplication in A × B is coordinate-wise.

E XERCISE 3. We find conjugacy class decompositions.

(a) Let G = Z2 ×S 3 . Write Z2 = {±1}. Then the conjugacy classes of Z2 are {1} and {−1}.
The conjugacy classes in S 3 are {1}, {(12), (13), (23)}, and {(123), (132)}. Hence
by the above note, the conjugacy classes in G are

{1} × {1} {1} × {(12), (13), (23)} {1} × {(123), (132)}

{−1} × {1} {−1} × {(12), (13), (23)} {−1} × {(123), (132)}

(b) Let G = S 3 × S 3 . The conjugacy classes in G are just the direct products of the
conjugacy classes in S 3 with each other.
(c) Let G = Z3 × A 4 . The conjugacy classes in G are just the direct products of the
conjugacy classes in Z3 ({1}, {x}, and {x 2 }) with the conjugacy classes in A 4 .

44
E XERCISE 5. Let G be finite and suppose |G : Z (G)| = n. Then every conjugacy class
in G has at most n elements.

Proof. Let g ∈ G. Note that Z (G) ≤ CG (g ) ≤ G, hence

n = |G : Z (G)| = |G : CG (g )||CG (g ) : Z (G)| ≥ |G : CG (g )|

Therefore by the Orbit-Stabilizer Theorem, the conjugacy class containing g has at

most n elements. Since g was arbitrary this establishes the claim.

E XERCISE 6. Let G be a nonabelian group of order 15. Then there is only one possible
class equation for G.

Proof. We must have Z (G) = 1, since otherwise G/Z (G) is of prime order and hence
cyclic, contradicting that G is nonabelian. Thus there is only the trivial singleton
conjugacy class. The other conjugacy classes must have orders 3 or 5. It is easy to
check that the only possibility is 15 = 1 + 3 + 3 + 3 + 5.

E XERCISE 13. Let G be a finite group with two conjugacy classes. Then G ∼
= Z2 .

Proof. Let n = |G|. Then we must have n = 1+k where k is the order of the nontrivial
conjugacy class. But k|n by the Orbit-Stabilizer Theorem. Hence there exists m with
mk = n = k + 1. Now m ≥ 1, but since k ≥ 1, m 6= 1. Also 2k > 1, so 3k > k + 1, and
hence we have m < 3. This leaves m = 2, in which case k = 1, n = 2, and G ∼
= Z2 .

E XERCISE 17. Let A be a nonempty set and X ⊆ S A . Then

F (X ) = { a ∈ A | (∀σ ∈ X )(σ(a) = a) }

is called the fixed set of X and M (X ) = A − F (X ) is called the moved set of X .

Set D = { σ ∈ S A | M (σ) < ∞ }. Then D E S A .

Proof. Trivially 1 ∈ D. Suppose σ, τ ∈ D, so that M (σ) and M (τ) are finite. Note that
M (τ−1 ) = τ[M (τ)] = M (τ), and M (στ−1 ) ⊆ M (σ) ∪ M (τ−1 ), hence M (στ−1 ) is finite
and στ−1 ∈ D. This shows that D ≤ S A .
Suppose now σ ∈ D and τ ∈ S A . Note that M (τστ−1 ) = τ[M (σ)], hence

|M (τστ−1 )| = |τ[M (σ)]| = |M (σ)| < ∞

since τ is a permutation. Thus τστ−1 ∈ D. Since σ, τ were arbitrary, D E S A .

E XERCISE 18. Let A be a nonempty set, H ≤ S A , and F (H ) the fixed set of H . Then
NS A (H ) stabilizes F (H ) and its complement M (H ).

Proof. Suppose τ ∈ N (H ). We claim that τ[F (H )] = F (H ). Note that τH = H τ, hence

if a ∈ F (H ) and σ ∈ H , there exists σ0 ∈ H such that

σ(τ(a)) = (στ)(a) = (τσ0 )(a) = τ(σ0 (a)) = τ(a)

Thus τ(a) ∈ F (H ). Since a was arbitrary, τ[F (H )] ⊆ F (H ).

By symmetry, τ−1 [F (H )] ⊆ F (H ), so F (H ) ⊆ τ[F (H )], and hence τ[F (H )] = F (H )
as claimed. Since τ is a bijection, it follows that τ[M (H )] = M (H ) as well.

45
E XERCISE 19. Let G be a group, H E G, and K be a conjugacy class of G such that
K ⊆ H . Then K is a union of k conjugacy classes of equal size in H , where for any
x ∈ K , k = |G : HCG (x)|.

Proof. This follows from the proof of Exercise 1.9. Note that under conjugation,
G acts transitively on K , hence K is partitioned into H -orbits of equal cardinality.
Now G acts transitively on the set of H -orbits. Hence k is just the size of the one
orbit under this action. The stabilizer of this action is precisely HCG (x), hence by
the Orbit-Stabilizer Theorem we have k = |G : HCG (x)| as desired.

Recall A n E S n and |S n | = 2·|A n | (hence A n is maximal for n ≥ 2). It follows from this
result that any conjugacy class in S n which is contained in A n consists either of one
conjugacy class in A n or else two conjugacy classes in A n of equal size.

E XERCISE 22. Suppose n ≥ 3 is odd. Then the set of all n-cycles forms two conjugacy
classes of equal size in A n .

Proof. Since n is odd, the n-cycles are even and hence are contained in A n . Define
σ = (123 · · · n), τ = (12), and σ∗ = τστ−1 = (213 · · · n). We claim that σ and σ∗ are
not conjugate in A n . Indeed, if ρ ∈ S n and ρσρ −1 = σ∗ , then by Proposition 10 and
the fact that the only alternate representations of the cycle for σ∗ are obtained by
cyclically permuting its elements, we must have ρ = (σ∗ )k τ for some k, which is an
odd permutation. Hence there is not one conjugacy class of n-cycles in A n , and the
desired result follows from Exercise 19.

E XERCISE 25. Let G = GL 2 (C) and let H be the subgroup of upper triangular matrices
in G. Then every element in G is conjugate to some element in H .

Proof. Let A be a matrix in G and let T be the corresponding linear transformation

under the standard basis. We know T has an eigenvalue because its characteristic
polynomial must have a complex root. By changing to a basis whose first element is
an eigenvector of T , we obtain an upper triangular matrix representation A ∗ for T .
But A ∗ is obtained by conjugating A, hence A is a conjugate of A ∗ . Since A was
arbitrary, this completes the proof.

E XERCISE 29. Let G be a group of order p α . Then G has a subgroup of order p β for
all 0 ≤ β ≤ α.

Proof. We proceed by induction on α. For α = 0 and α = 1 the result is trivial.

Suppose α > 1 and the result holds for all groups of order p γ with γ < α. By
Theorem 8, Z (G) 6= 1, hence |Z (G)| = p δ for 0 < δ ≤ α. Since Z (G) is normal, we may
define G = G/Z (G). Then |G| = p α−δ by Lagrange’s Theorem and α − δ < α, so by the
induction hypothesis we may assume the result holds for G.
Now let 0 ≤ β ≤ α be arbitrary. If β ≤ δ, then p β |p δ , hence by Exercise 4.4 in
Chapter 3 there exists a subgroup of Z (G) of order p β . If β > δ, then ρ = β−δ satisfies
0 < ρ ≤ α − δ, so there exists a subgroup H of G with |H | = p ρ . But we know by the
Lattice Theorem that H = H /Z for some subgroup Z E H ≤ G, and hence

|H | = |H ||Z | = p ρ p δ = p β−δ p δ = p β

46
Thus in either case there exists a subgroup of G of order p β , as desired.

E XERCISE 30. Let G be a group of odd order. Then for all nonidentity elements x ∈ G,
x is not conjugate to x −1 .

Proof. Suppose x ∈ G, x 6= 1 and x is conjugate to x −1 . We cannot have x = x −1 ,

lest x 2 = 1 so x has order 2 and 2 divides |G| by Lagrange’s Theorem—contradicting
that |G| is odd. We claim that x has an even number of conjugates. Indeed, each
element axa −1 in the conjugacy class of x can be paired off with its distinct inverse
element (axa −1 )−1 = ax −1 a −1 , which is also in the conjugacy class of x. But then by
the Orbit-Stabilizer Theorem, the order of the conjugacy class divides the order of G,
so 2 divides |G|—again a contradiction. It follows that our initial assumption cannot
be correct, so there is no nonidentity element conjugate to its inverse.

E XERCISE 33. Let σ ∈ S n , let m 1 , . . . , m s be the distinct cycle lengths in the (complete)
cycle decomposition of σ, and for each 1 ≤ i ≤ s let k i be the number of m i -cycles in
the decomposition (so n = is=1 k i m i ). Then the number of conjugates of σ is
P

s
Y 1
n! k
i =1 ki ! mi i

Proof. Recall that the number of conjugates of σ is precisely the number of elements
in S n with the same cycle type as σ, hence we need only count the number of ways
to form a permutation in S n with this cycle type.
We give a combinatorial argument. To obtain any such permutation we may start
with any linear arrangement of the numbers 1, . . . , n and then partition the numbers
into cycles, by drawing parentheses, according to the cycle type of σ. Now there
are n! distinct linear arrangements of the numbers 1, . . . , n. However, for any such
permutation formed, for each 1 ≤ i ≤ s, the m i -cycles may be permuted amongst
each other in any of k i ! ways, and the elements in each m i -cycle may be cyclically
permuted in m i ways, while still preserving identity and form. Thus each linear
k
arrangement of 1, . . . , n is in an equivalence class of is=1 k i ! m i i (including itself),
Q

all of which give the same permutation. Thus the number of distinct permutations
with the cycle type of σ is as above.

E XERCISE 36. Let G be a group. Let π : G → SG be the left regular representation of G,

and write π(g ) = σg , so σg (x) = g x for all x ∈ G. Similarly let λ : G → SG be the right
regular representation of G, and write λ(h) = τh , so τh (x) = xh −1 for all h ∈ G.1

(a) C SG (π(G)) = λ(G) and C SG (λ(G)) = π(G). In particular, σg and τh commute for
all g , h ∈ G.
1 Note that under our definitions, the right regular representation λ of G is not associated with a right
action of G, but a left action h · x = xh −1 , for otherwise λ would fail to be a homomorphism under our
usual left-based functional notation.

47
 Proof. We prove the first equality; the other follows by symmetry. Note that
λ(G) ⊆ C SG (π(G)) since for g , h ∈ G,

(τh σg )(x) = τh (σg (x)) = (g x)h −1 = g (xh −1 ) = σg (τh (x)) = (σg τh )(x)

for all x ∈ G. Conversely, suppose τ ∈ C SG (π(G)), so τσg = σg τ for all g ∈ G. We

claim that there exists some h ∈ G such that τ = τh . Indeed, let h = τ(1)−1 , so
τ(1) = h −1 . Then

τ(x) = τ(σx (1)) = (τσx )(1) = (σx τ)(1) = σx (τ(1)) = σx (h −1 ) = xh −1 = τh (x)

for all x ∈ G. Hence τ = τh , satisfying the claim.

Thus C SG (π(G)) = λ(G) as desired.

(b) σg = τg iff g ∈ Z (G) and |g | equals 1 or 2. Proof: part (c).

(c) σg = τh iff g , h ∈ Z (G) and g = h −1 . Thus π(G) ∩ λ(G) = π(Z (G)) = λ(Z (G)).

Proof. Note σg = τh iff for all x ∈ G, g x = σg (x) = τh (x) = xh −1 , which is true

iff g = h −1 (taking x = 1) and hence g x = xg for all x ∈ G, which is true iff
g = h −1 and g , h ∈ Z (G).
The set identities are now immediate.

Note this exercise shows in particular that the centralizer of the left or right regular
permutation representation of a group is isomorphic to the group itself.

Section 4

E XERCISE 1. Let G be a group, g ∈ G, and let ϕg denote conjugation by g . Then for

σ ∈ Aut(G), σϕg σ−1 = ϕσ(g ) . Hence Inn(G) E Aut(G).

Proof. Let ϕ∗ = σϕg σ−1 . Then for x ∈ G,

ϕ∗ (x) = (σϕg σ−1 )(x) = σ(g σ−1 (x)g −1 ) = σ(g )xσ(g )−1

Hence ϕ∗ = ϕσ(g ) as desired.

The quotient Aut(G)/Inn(G) is called the outer automorphism group of G.

48
E XERCISE 2. Let G be an abelian group, |G| = pq where p 6= q are prime. Then G is
cyclic.

Proof. By Cauchy’s Theorem, there exist x, y ∈ G with |x| = p and |y| = q. Set z = x y.
Note that
z pq = (x y)pq = x pq y pq = (x p )q (y q )p = 1 · 1 = 1
Hence |z| divides pq. We claim |z| = pq.
If |z| divides p, then 1 = (x y)p = x p · y p = y p . But this implies q|p, which is a
contradiction since p 6= q. Similarly |z| does not divide q since p does not divide q.
Hence |z| = pq as claimed, and G is cyclic.

Recall from the text that if |G| = pq with p ≤ q and p does not divide q − 1, then G is
abelian. Combining this with the above result we obtain the following classification
theorem: if |G| = pq with p < q and p does not divide q − 1, then G is cyclic.
As an example, if |G| = 15 = 3 · 5, then G is cyclic.

E XERCISE 3. |Aut(D 8 )| ≤ 8

Proof. Looking at cycle decompositions (Exercise 1.7.11) of the elements in D 8 , we

see that under any automorphism, r has at most two images (elements of order 4),
and s has at most four images (elements of order 2 which are not in the center).

E XERCISE 5. Aut(D 8 ) ∼
= D8

Proof. Note D 8 E D 16 , so D 16 acts on D 8 by conjugation and D 16 /C D 16 (D 8 ) is iso-

morphic to a subgroup of Aut(D 8 ). Note that C D 16 (D 8 ) = 〈r 4 〉, so since |Aut(D 8 )| ≤ 8
(by Exercise 3), we have
Aut(D 8 ) ∼
= D 16 /〈r 〉 ∼
4
= D8
where the last isomorphism holds since the groups have an identical presentation.

This exercise illustrates that if inner automorphisms do not exhaust Aut(G), then it
is useful to look at larger normalizers of G.

E XERCISE 6. If H char G, then H E G.

Proof. Trivial since H is preserved under inner automorphisms.

E XERCISE 7. If H is the unique subgroup of a given order in G, then H char G.

Proof. Trivial since automorphisms preserve order and map subgroups to subgroups.

Note it follows that any subgroup of a cyclic group is characteristic.

49
E XERCISE 8. Suppose H ≤ K ≤ G.

(a) If H char K E G, then H E G.

Proof. If g ∈ G and ϕg denotes the automorphism in G of conjugation by g ,

then ϕg |K ∈ Aut(K ) since K E G. Hence ϕg (H ) = ϕg |K (H ) = H since H char K ,
that is, g H g −1 = H . Since g was arbitrary, H E G.

(b) If H char K char G, then H char G.

Proof. If σ ∈ Aut(G), then σ(K ) = K since K char G, hence σ|K ∈ Aut(K ). But
then σ(H ) = σ|K (H ) = H since H char K . Since σ was arbitrary, H char G.

Note that V4 char S 4 .

(c) It is not true in general that H E K char G implies H E G. For example, Z2 E
V4 char S 4 , but Z2 is not normal in S 4 .

E XERCISE 9. Let G = D 2n and K = 〈r 〉. Then every subgroup of K is normal in G.

Proof. Let H ≤ K . Note K E G (by checking generators). Recall that Aut(K ) consists
of maps r 7→ r a where 0 < a < n and (a, n) = 1. Hence H is mapped to itself under any
automorphism of K , that is, H char K . It follows from Exercise 8(a) that H E G.

This exercise illustrates another useful way of finding normal subgroups.

E XERCISE 11. Let G = S p , and P ≤ G with |P | = p. Then NG (P )/CG (P ) ∼

= Aut(P ).

Proof. P is cyclic and generated by a p-cycle x. We know Aut(P ) consists of maps

x 7→ x k with 0 < k < p, where the image of x is always a p-cycle. Each of these
maps can be obtained via conjugation of x in S p w ith an element of N (P ). Hence
N (P )/C (P ) is isomorphic to all of Aut(P ).

E XERCISE 13. If |G| = 203 and G has a normal subgroup of order 7, then G is abelian
(and hence cyclic).

Proof. Note |G| = 7 · 29, where 7 and 29 are both prime. Let H E G with |H | = 7. We
know G/C (H ) is isomorphic to a subgroup of Aut(H ), hence |G|/|C (H )| must divide
|Aut(H )| = ϕ(7) = 6. This is only possible if C (H ) = G, that is, H ≤ Z (G). This implies
G/Z is of prime order and hence cyclic, so G is abelian. It follows that G is cyclic by
Exercise 2.

E XERCISE 15. The groups (Z/5Z)× , (Z/9Z)× , and (Z/18Z)× are each cyclic.

Proof. The groups are each abelian. The latter two are of orders ϕ(9) = 3(3 − 1) = 6
and ϕ(18) = (2 − 1) · 3(3 − 1) = 6, respectively, hence cyclic by Exercise 2. The first
group is cyclic and generated by the element 2.

Section 5

50
N OTE .The proof of Sylow’s Theorem illustrates several recurring techniques:
1. It uses induction to prove (part of) an existence claim (cf. Cauchy’s Theorem
for abelian groups, Exercise 3.29, etc.).
2. It makes use of the Class Equation, combining information about the order
of G and an assumption about the order of Z (G) to obtain information about
subgroups (centralizers) in G (cf. an example in Section 4.4).
3. To count Sylow subgroups, it restricts the source of a group action to a sub-
group about which we have more information, which provides additional in-
formation about the original orbits (cf. Exercise 1.9).
4. In using Sylow p-subgroups computationally, it utilizes their maximality in
order among p-subgroups.
These techniques are important to keep in mind.

E XERCISE 3 (C AUCHY ’ S T HEOREM ). Let G be a group, p a prime dividing |G|. Then

there exists g ∈ G with |g | = p.

Proof. Let P be a Sylow p-subgroup of G. Since p divides |G|, there exists x ∈ P ,

x 6= 1. Set H = 〈x〉. Then H is a nontrivial cyclic p-subgroup of G, so p divides |H |
and hence there exists g ∈ H with |g | = p.

E XERCISE 5. Let G = D 2n . If p is an odd prime and P ∈ Sylp (G), then P is cyclic and
normal.

Proof. Let P ∈ Sylp (G). Let r, s be the usual generators of G and set H = 〈r 〉. Recall
|H | = n, hence (since p 6= 2) any Sylow p-subgroup of H is also a Sylow p-subgroup
of G. By Sylow’s Theorem then, all Sylow p-subgroups of G are contained in H , so
in particular P ≤ H . It follows that P is cyclic, and P char H E G (see Exercise 4.7),
hence P E G.

E XERCISE 6. We find all Sylow 3-subgroups of S 4 and A 4 .

Proof. Recall |S 4 | = 4! = 24 = 23 ·3, hence a Sylow 3-subgroup of S 4 is just a subgroup

of order 3 generated by an element of order 3. By direct computation it is easily seen
that the distinct such subgroups are:
〈(123)〉 〈(124)〉 〈(134)〉 〈(234)〉
Since these consist of even permutations, they are also the Sylow 3-subgroups of A 4 .

E XERCISE 13. Let G be a group with |G| = 56. Then G has a normal Sylow p-subgroup
for some prime p dividing |G|.

Proof. Note |G| = 23 ·7. By Sylow’s Theorem, n 7 ≡ 1 (7) and n 7 |8, thus n 7 = 1 or n 7 = 8.
If n 7 = 1, then there is a unique normal Sylow 7-subgroup of G. If n 7 = 8, then there
are 8 · (7 − 1) = 48 elements of order 7 in G (since distinct Sylow 7-subgroups must
have trivial intersection), leaving 8 remaining elements. These 8 elements must be
precisely those in a unique, and hence normal, Sylow 2-subgroup of G.

51
E XERCISE 16. Let G be a group with |G| = pqr , where p, q, r are prime and p < q < r .
Then G has a normal Sylow subgroup for one of p, q, r .

Proof. Assume towards a contradiction that none of n p , n q , n r equal 1. We know

from Sylow’s Theorem that n p |qr , hence n p ≥ q. Also n q |pr , but since n q ≡ 1 (q)
and p < q, we must have n q ≥ r . Similarly n r |pq but n r 6= p and n r 6= q, so n r = pq.
But then by counting elements we must have

pqr = |G| ≥ q(p − 1) + r (q − 1) + pq(r − 1)

= pqr + qr − (q + r )
> pqr + qr − 2r since q < r
> pqr since 2 ≤ p < q

—a contradiction. Thus one of n p , n q , n r equals 1, establishing the result.

E XERCISE 17. Let G be a group with |G| = 105. Then there exists a normal Sylow
5-subgroup and a normal Sylow 7-subgroup in G.

Proof. Note |G| = 3 · 5 · 7. By Sylow’s Theorem, n 7 ≡ 1 (7) and n 7 |15, hence n 7 = 1 or

n 7 = 15. Similarly n 5 = 1 or n 5 = 21. If neither n 5 nor n 7 equals 1, then there exist
15 · (7 − 1) + 21 · (5 − 1) = 174 > 105 elements of orders 7 and 5 in G, contradicting the
order of G. Thus at least one of n 7 and n 5 equals 1.
Let P ∈ Syl7 (G) and Q ∈ Syl5 (G). We know from the above remarks that P E G or
Q E G, so PQ ≤ G. Now |P ∩Q| = 1, hence |PQ| = |P | |Q| = 7 · 5 = 35. It follows that
PQ is cyclic (since 5 < 7 and 5 does not divide 7 − 1), and PQ E G since |G : PQ| = 3
(the smallest prime dividing |G|). Hence any subgroup of PQ is normal in G (see
Exercises 4.7–8), so in particular P E G and Q E G.

Note this proof utilizes the technique of constructing an intermediate cyclic normal
subgroup (cf. the example on groups of order 30).

E XERCISE 22. Let G be a group and |G| = 132. Then G is not simple.

Proof. Note |G| = 22 ·3·11. By Sylow’s Theorem, n 11 = 1 or n 11 = 12. If n 11 = 1, G has a

normal Sylow 11-subgroup and we are done. If n 11 = 12, there exist 12·(11−1) = 120
elements of order 11 in G, leaving 12 elements remaining.
In this case, if n 3 = 1 there exists a normal Sylow 3-subgroup and we are done.
Otherwise we must have n 3 = 4 and 4 · (3 − 1) = 8 elements of order 3, which leaves
4 elements in G to form the unique and hence normal Sylow 2-subgroup of G.

52
E XERCISE 24. Let G be a group, |G| = 231. Then Z (G) contains a Sylow 11-subgroup
and there exists a normal Sylow 7-subgroup.

Proof. Note |G| = 3 · 7 · 11. By Sylow’s Theorem, n 7 ≡ 1 (7) and n 7 |33, leaving only the
possibility n 7 = 1 and a normal Sylow 7-subgroup.
Similarly n 11 = 1. Let P be the unique Sylow 11-subgroup of G. Since P is normal,
N (P ) = G, and hence G/C (P ) is isomorphic to a subgroup of Aut(P ). In particular,
|G|/|C (G)| divides |Aut(P )| = ϕ(11) = 10. But this is only possible if |C (P )| = |G|, that
is, C (P ) = G so P ≤ Z (G).

E XERCISE 26. Let G be a group with |G| = 105 and suppose G has a normal Sylow
3-subgroup. Then G is abelian.

Proof. Recall |G| = 3 · 5 · 7. Let P ∈ Syl3 (G). Then N (P ) = G, so G/C (P ) is isomorphic

to a subgroup of Aut(P ). In particular, |G|/|C (P )| divides |Aut(P )| = ϕ(3) = 2. But this
is only possible if C (P ) = G, that is, P ≤ Z (G).
Set G = G/Z (G). The only possibilities for |G| are 1, 5, 7, 35(= 5 · 7). In all cases,
G is cyclic (for case 35, note 5 < 7 and 5 does not divide 7−1). Hence G is abelian.

E XERCISE 30. A simple group of order 168 must have 48 elements of order 7.

Proof. Suppose |G| = 168 = 23 · 3 · 7. By Sylow’s Theorem, n 7 ≡ 1 (7) and n 7 |24, hence
n 7 = 1 or n 7 = 8. But since G is simple, n 7 6= 1. Hence since every element of order 7
lies in a Sylow 7-subgroup, and distinct Sylow 7-subgroups have trivial intersection,
there are 8 · (7 − 1) = 48 elements of order 7 in G.

E XERCISE 32. Suppose P ∈ Sylp (H ) and P E H E K . Then P E K .

Proof. By Corollary 20, P char H E K , hence P E K by Exercise 4.8(a).

Set H = N (P ), so P E H E N (H ). By this exercise, P E N (H ), hence N (H ) ≤ H , so

N (H ) = H . In other words, normalizers of Sylow p-subgroups are self-normalizing.

E XERCISE 33. Let G be a group, P a normal Sylow p-subgroup of G, and H ≤ G. Then

P ∩ H is the unique Sylow p-subgroup of H .

Proof. Let P ∗ be a Sylow p-subgroup of H . Then since P ∗ is a p-subgroup of G, there

exists g ∈ G with P ∗ ≤ g P g −1 = P . Hence P ∗ ≤ P ∩ H . Now P ∩ H is a p-subgroup
of H , so by the maximality of P ∗ we have P ∗ = P ∩ H . Since P ∗ was arbitrary, P ∩ H
is unique. (Alternately, this follows since P ∩ H E H .)

53
E XERCISE 34. Let G be a group, P a Sylow p-subgroup of G, and N E G. Then P ∩ N
is a Sylow p-subgroup of N .

Proof. Let P ∗ ∈ Sylp (N ). Then there exists g ∈ G such that P ∗ ≤ g P g −1 , and hence

g −1 P ∗ g ≤ g −1 (g P g −1 )g = P

Set P ∗∗ = g −1 P ∗ g . Then P ∗∗ ∈ Sylp (N ) since N E G, and by counting elements we

get P ∗∗ = P ∩ N (cf. Exercise 33).

Note that
|P N | |P ||N | |P |
|P N /N | = = =
|N | |P ∩ N ||N | |P ∩ N |
which is the highest power of p dividing |G|/|N |. Hence P N /N ∈ Sylp (G/N ). (See
also Exercise 3.3.9.)

E XERCISE 35. Note that for P ∈ Sylp (G) and H ≤ G, it is not true in general that
P ∩H ∈ Sylp (H ). For example, P = 〈(12)〉 is a Sylow 2-subgroup of S 3 , but if H = 〈(23)〉,
P ∩ H = 1, which is not a Sylow 2-subgroup of H .

E XERCISE 43. Let G be the group of rigid motions in R3 of a (regular) dodecahedron.

Then G ∼= A5.

Proof. We know from a counting argument that |G| = 60 (see Exercise 1.2.12). Hence
by Proposition 23, it suffices to prove that G is simple, and hence by Proposition 21
it suffices to prove that G has more than one Sylow 5-subgroup.
Since 60 = 22 · 3 · 5, any subgroup of G of order 5 is a Sylow 5-subgroup of G.
Recall that G acts on the set of 12 faces of the dodecahedron and each such face has
a stabilizer of order 5. Clearly there exists more than one distinct such stabilizer. By
the above remarks this completes the proof.

E XERCISE 44. Let G be a group and p be the smallest prime dividing |G|. Then if
P ∈ Sylp (G) and P is cyclic, N (P ) = C (P ).

Proof. Write |P | = p α . Since P is cyclic, P ≤ C (P ) ≤ N (P ). Let k = |N (P )|/|C (P )|.

We know k divides |Aut(P )| = ϕ(p α ) = p α−1 (p − 1). But since p α divides |C (P )| this
implies k divides p − 1. Also, since p is the smallest prime dividing |G|, no nontrivial
factor of |N (P )| divides p − 1, so k = 1. This implies N (P ) = C (P ) as desired.

E XERCISE 49. Let G be a group of order 2n m, where m is odd, and suppose G has a
cyclic Sylow 2-subgroup. Then G has a normal subgroup of order m.

Proof. We prove existence of a subgroup of order m by induction on n.

Case n = 0 is trivial. If n > 0, since G has a cyclic Sylow 2-subgroup, we may fix
x ∈ G with |x| = 2n . Then |x| is even and |G|/|x| is odd. Hence by Exercises 2.11–12,
there exists H ≤ G with |G : H | = 2, that is, |H | = 2n−1 m. By induction there exists
K E H with |K | = m.

Section 6

54
E XERCISE 1. For n ≥ 5, A n has no proper subgroups with index strictly less than n.

Proof. Suppose towards a contradiction that A < A n and |A n : A| = m < n.

Recall A n acts on the coset space A n /A, affording a permutation representation
ϕ : A n → S m . Write K = ker ϕ. Then K ≤ A, and A n /K is isomorphic to a subgroup
of S m . Since 1 < m < n, we know |S m | = m! < n!/2 = |A n |, hence K 6= 1. But then
1 < K ≤ A < A n and K E A n (since K is a kernel), so K is a nontrivial proper normal
subgroup of A n . This contradicts that A n is simple for n ≥ 5.
S
E XERCISE 5. Let G be a group and suppose G = G i where G 1 ≤ G 2 ≤ · · · ≤ G and
each G i is simple. Then G is simple.

Proof. Suppose towards a contradiction G is not simple, so there exists 1 < H < G
with H E G. For all i , set Hi = G i ∩ H . Then H1 ≤ H2 ≤ · · · ≤ H = Hi and Hi E G i
S

for all i . Since H 6= 1, there exists a least j with H j 6= 1. If Hk = G k for all k ≥ j , then
[ [ [
H= Hi = Hk = Gk = G
k≥ j k≥ j

contradicting that H < G. Hence there exists k ≥ j such that 1 < H j ≤ Hk < G k —
in other words, Hk is a nontrivial proper normal subgroup of G k , contradicting our
assumption that G k is simple.

Chapter 5
Section 1
Q
E XERCISE 4. Let G 1 , · · · ,G n be finite groups, G = G i , and p a prime. Then
©Y ª
Sylp (G) = P i | P i ∈ Sylp (G i )
Q
Hence n p (G) = n p (G i ).
Q Q Q
Proof. The reverse inclusion is immediate since |G| = |G i | and | P i | = |P i | for
any subgroups P i ≤ G i .
To prove the forward inclusion, suppose P ∈ Sylp (G), and set P i = P ∩ G i for all i
Q
so P = P i (here we are identifying G i with the naturally corresponding subgroup
of G). Then by Lagrange’s Theorem, P i is a p-subgroup of G i for all i , so there exist
subgroups P i∗ ∈ Sylp (G i ) with P i ≤ P i∗ . Note P ∗ = P i∗ ∈ Sylp (G) by the above, and
Q
∗ ∗ ∗ ∗
P ≤ P . But P = P since |P | = |P |, hence P i = P i and P i ∈ Sylp (G i ) for all i .
The value of n p (G) follows immediately.

55
E XERCISE 5. We exhibit a nonnormal subgroup of G = Q 8 × Z4 . Recall

Q 8 = {±1, ±i , ± j , ±k} where i 2 = j 2 = k 2 = i j k = −1

Write Z4 = 〈x〉. Set H = 〈i x〉 = {1, i x, −x 2 , −i x 3 }. Then H is not normal in G, since e.g.

j (i x) j −1 = j (i x)(− j ) = j i (− j )x = j (−k)x = −i x 6∈ H

Note this exercise illustrates the importance of looking at subgroups of products

other than the factors in the product. Indeed, all subgroups of Q 8 and Z4 are normal,
hence normal in G, and hence any product obtained from these subgroups is also
normal in G. The subgroup H is isomorphic to Z4 but is not normal in G.

E XERCISE 7. Let G 1 , · · · ,G n be groups and π ∈ S n . Define

ϕπ : G i → G π−1 (i )
Y Y

(g i ) 7→ (g π−1 (i ) )

Then ϕπ is an isomorphism.

Proof. Clearly ϕπ is well defined. Also

ϕπ ((g i )(h i )) = ϕπ (g i h i )
= (g π−1 (i ) h π−1 (i ) )
= (g π−1 (i ) )(h π−1 (i ) )
= ϕπ (g i )ϕπ (h i )

So ϕπ is a homomorphism. If (g π−1 (i ) ) = (h π−1 (i ) ), then g i = g π(π−1 (i )) = h π(π−1 (i )) = h i

for all i , that is, (g i ) = (h i ). If (g i∗ ) ∈ G π−1 (i ) , set (g i ) = (g π(i
∗
Q Q
)
). Then (g i ) ∈ G i and
∗ ∗
ϕπ (g i ) = (g π−1 (i ) ) = (g π(π −1 (i )) ) = (g i )

Hence ϕπ is a bijection, so ϕπ is an isomorphism as desired.

E XERCISE 8. Let G 1 = · · · = G n and G = G i . Then for π ∈ S n and ϕπ as above,

Q

ϕπ ∈ Aut(G), and the map π 7→ ϕπ is an embedding of S n into Aut(G).

Proof. It follows from Exercise 7 that ϕπ ∈ Aut(G). If σ, τ ∈ S n , then

(ϕσ ◦ ϕτ )(g i ) = ϕσ (ϕτ (g i ))

= ϕσ (g τ−1 (i ) )
= (g τ−1 (σ−1 (i )) )
= (g (στ)−1 (i ) ) = ϕστ (g i )

(Note the third equality follows since if we write h i = g τ−1 (i ) , then ϕσ (h i ) = (h σ−1 (i ) )
by definition and h σ−1 (i ) = g τ−1 (σ−1 (i )) .) Therefore π 7→ ϕπ is a homomorphism. It is
clear that the kernel of this homomorphism is trivial, so it is injective.

56
E XERCISE 9. Let F be a field, and let H be the set of n × n matrices over F with only a
single 1 in each row and column (and 0’s elsewhere). Then H ≤ GL n (F ) and H ∼ = Sn .

Proof. Note H acts faithfully on the standard basis vectors e 1 , . . . , e n in the vector
space F n . Hence H ≤ GL n (F ) (since each matrix in H has a natural inverse), and
H∼= Sn .

The matrices in H are called permutation matrices for the obvious reason.

E XERCISE 11. Let n be a positive integer and p a prime. The number of subgroups
of order p in the elementary abelian group E p n is (p n − 1)/(p − 1).

Proof. Each of the p n − 1 nonidentity elements in E p n generates, and hence lives in,
a subgroup of order p. By Lagrange’s Theorem, distinct subgroups of order p have
trivial intersection, hence the p n − 1 nonidentity elements are partitioned into sets
of p − 1 elements, namely the subgroups of order p minus the identity element. It
follows that there are (p n − 1)/(p − 1) such subgroups.

E XERCISE 12,13 (C ENTRAL P RODUCTS ). Let A and B be groups with Z1 ≤ Z (A) and
Z2 ≤ Z (B ) such that Z1 ∼
= Z2 by the mapping x i 7→ y i . Set Z = { (x i , y i−1 ) | x i ∈ Z1 } and
define A ∗ B = (A × B )/Z .

(a) Let A = A/Z and B = B /Z (here identifying A and B with A × 1 and 1 × B ,

respectively). Then A ∼
= A and B ∼
= B , and A ∩ B is a central subgroup of A ∗ B
isomorphic to Z1 ( and Z2 ).

Proof. Let π project A onto A/Z . Then ker π = {(a, 1) ∈ Z }. But if (a, 1) ∈ Z ,
(a, 1) = (x i , y i−1 ) for some i , that is, y i−1 = 1 so y i = 1 and hence a = x i = 1.
Thus ker π = 1 and A ∼ = A by the First Isomorphism Theorem. Similarly B ∼ = B.
If x ∈ A ∩ B , then x = a = b for some a ∈ A, b ∈ B . In particular, (a, 1) ∈ (1, b)Z ,
so there exists i with

(a, 1) = (1, b)(x i , y i−1 ) = (x i , b y i−1 )

That is, a = x i and b = y i , so x = x i = y i . Conversely, any element of the

form x i is in A ∩ B , since x i = y i in A ∗ B . Hence we have

A ∩ B = { x i | x i ∈ Z1 } = Z1 = Z2 = { y i | y i ∈ Z2 }

which is isomorphic to Z1 (and Z2 ). This subgroup is central in A ∗ B since

Z1 and Z2 are central in A and B , respectively.

(b) Let Z4 = 〈x〉 and D 8 = 〈r, s〉 (r, s as usual). Then we may form Z4 ∗D 8 under the
identification x 2 7→ r 2 . We then have

Z4 ∗ D 8 = 〈x, r , s | x 4 = 1 = r 4 = s 2 , r s = s r −1 , x 2 = r 2 〉

57
The central product illustrates well how quotient groups can be used to ‘identify’
elements in existing groups. In this case we desire to identify each element x i in
a central subgroup of A with a corresponding element y i in an isomorphic central
subgroup of B . In other words, we desire x i = y i . We achieve this by taking a quotient
to force x i = y i . Since

x i = y i ⇐⇒ x i y i −1 = 1 ⇐⇒ x i y i−1 = 1

we divide out the elements x i y i−1 = (x i , y i−1 )—that is, the elements in Z .
In the quotient A ∗ B , we retain the structures of A and B , but the two distinct
central subgroups of A and B have collapsed into one central subgroup of both.

E XERCISE 14. Suppose B i E A i for all i . Then B i E A i and A i / B i ∼

Q Q Q Q Q
= (A i /B i )

Proof. Define π : A i → (A i /B i ) by the map (a i ) 7→ (a i B i ). It is immediate that π is

Q Q

a surjective homomorphism and ker π = B i . Hence B i E A i and the quotient

Q Q Q

identity follows from the First Isomorphism Theorem.

Note that even if every subgroup of each A i is normal, it is not necessarily true that
Q
every subgroup of A i is normal (see Exercise 5).

E XERCISE 18. We exemplify groups satisfying various properties:

(a) G = ∞
Q
n=1 Z2 is an infinite group, but every element has order 1 or 2.
(b) G = Q/Z is an infinite group in which every element has finite order, but there
is an element of order n for each positive integer n (cf Exercise 3.1.14).
(c) G = Z2 × Z has an element of order 2 and an element of infinite order.
(d) G = ∞
Q
n=1 S n is such that any finite group is isomorphic to a subgroup of G.
(e) G =
Q∞
Z2 is nontrivial and G =∼ G ×G.
n=1

Section 2

E XERCISE 1. For each n, we count the number of isomorphism classes of abelian

groups of order n by counting the number of possible lists of elementary divisors.

(a) n = 100: Note n = 22 · 52 , hence there are 2 · 2 = 4 classes.

(b) n = 576: Note n = 26 · 32 , hence since there are 11 possible partitions of 6 ele-
ments, there are 11 · 2 = 22 classes.

E XERCISE 3. For each n, we find all possible elementary divisors and invariant fac-
tors for abelian groups of order n.

(a) n = 270: Note 270 = 2 · 33 · 5, hence the possible lists of elementary divisors are
{2, 33 , 5}, {2, 32 , 3, 5}, and {2, 3, 3, 3, 5}. Using the algorithm to obtain invariant
factors from elementary divisors, we obtain the corresponding lists {2 · 33 · 5},
{2 · 32 · 5, 3}, and {2 · 3 · 5, 3, 3}, respectively, of invariant factors.

58
E XERCISE 4. Let [a 1 , · · · , a k ] denote the group Z a1 × · · · × Z ak . We determine which
pairs of groups given are isomorphic.
(a) For groups [4, 9], [6, 6], [8, 3], [9, 4], [6, 4], and [64], we have the following lists of
elementary divisors:

[4, 9] −→ {22 , 32 }
[6, 6] −→ {2, 2, 3, 3}
[8, 3] −→ {23 , 3}
[9, 4] −→ {22 , 32 }
[6, 4] −→ {2, 22 , 3}
[64] −→ {26 }

Hence [4, 9] ∼
= [9, 4], but no other pairs of distinct groups above are isomorphic.
E XERCISE 5. Let G be a finite abelian group of type (n 1 , . . . , n t ). Then G contains an
element of order m iff m|n 1 . In particular, G has exponent n 1 .

Proof. By assumption G = Zn1 ×· · ·×Zn t . If m|n 1 , then there is an element of order m

in the subgroup Zn1 , and hence in G. Conversely, if x = (x 1 , . . . , x t ) ∈ G has order m,
then m = lcm(|x 1 |, . . . , |x t |). Now |x i | divides n i for 1 ≤ i ≤ t , hence since each n i
divides n 1 , n 1 is a common multiple of the |x i |. Since m is the least such multiple,
m|n 1 as desired.
It follows that for all x ∈ G, x n1 = 1 since |x| divides n 1 . Also G has an element of
order n 1 (namely the generator of Zn1 ), so no smaller number satisfies this property.
Therefore G has exponent n 1 .

E XERCISE 6. A finite group has finite exponent. An infinite group may have finite
exponent. A group with exponent m need not contain an element of order m.

Proof. If |G| = n, then for all x ∈ G, |x| divides n (Lagrange), hence x n = 1. Let m be
the least integer satisfying this property. Then m is the exponent of G.
Let G = ∞ Z2 . Then G is an infinite group with finite exponent 2.
Q

Let G = D 6 . Then G has exponent 6, but G contains no element of order 6.

E XERCISE 7. Let p be a prime. For 1 ≤ i ≤ n, let A i = 〈x i 〉 where |x i | = p αi > 1, and set

A = A i . Define ϕ : A → A by x 7→ x p (the p-th power map).
Q

(a) ϕ is a homomorphism. (Proof: trivial since each A i is abelian.)

Q p Q p αi −1
(b) imϕ = 〈x i 〉 and ker ϕ = 〈x i 〉.

m m m
Proof. Write [x i i ] = (x 1 1 , · · · , x n n ). Then
© m ª © m ª © p ª Y p
imϕ = [x i i ]p = [(x i i )p ] = [(x i )mi ] = 〈x i 〉
¡Q ¢p Q p
If we let A p denote imϕ on A, then this shows: A i = A i . Also
© m m
ker ϕ = [x i i ] | [x i i ]p = 1
ª

59
 mi p
= 1 for all i , which is true iff p αi |m i p for all i , which is
m
Now [x i i ]p = 1 iff x i
Q p αi −1
true iff p αi −1 |m i . Hence ker ϕ = 〈x i 〉 as desired.

(c) A/imϕ ∼
= Epn ∼
= ker ϕ, hence in particular A/imϕ and ker ϕ both have rank n.

Proof. By part (b), |ker ϕ| = ni=1 p = p n , and x p = 1 for all x ∈ ker ϕ, hence
Q

ker ϕ ∼
= E p n . Also
¡Y ¢ ¡Y p ¢ Y p
Ai ∼= (A i /A i ) ∼
Y
A/imϕ = Ai / = Zp = E p n
p
where the second isomorphism follows since |A i | = p αi −1 .

E XERCISE 8. Let p be a prime and G be a finite abelian group. Let G p and G p denote
the image and kernel of the p-th power map, respectively.

(a) G/G p ∼
= Gp .

Proof. Let n = |G|. If p does not divide n, the result is trivial since in this case
α α α
G p = G and G p = 1. Hence we assume p divides n. Write n = p 1 1 p 2 2 · · · p k k ,
α1
with pairwise distinct primes and p 1 = p (so p 1 > 1). By Theorem 5, we may
α Q p
write G = A i where |A i | = p i i for 1 ≤ i ≤ k. Now G p = A i , and it is easily
Q
p
seen that A i = A i and (A i )p = 1 for i > 1, hence we have
¡Y ¢ ¡Y p ¢
G/G p = Ai / Ai
∼
Y p
= (A i /A ) i
p
= (A 1 /A 1 ) × 1 × · · · × 1
∼
= (A 1 )p × 1 × · · · × 1 by Exercise 7
Y
= (A i )p
= Gp

Note we may appeal to Exercise 7 since A 1 is a nontrivial abelian p-group.

E XERCISE 13. Let A = ri=1 〈x i 〉 where |x i | = n i for 1 ≤ i ≤ r . Let G be any group with
Q
n
commuting elements g 1 , . . . , g r such that g i i = 1 for 1 ≤ i ≤ r . Then there is a unique
homomorphism from A to G which maps x i 7→ g i .
Q α Q α
Proof. Define ϕ : A → G by x i i 7→ g i i . Note ϕ is well defined since if x = y
for x, y ∈ A, then both x and y (as products of powers of the x i ) can be reduced,
using commutativity and the relations |x i | = n i , to the same expression of the form
Qr ki
i =1 x i where 0 ≤ k i < n i for 1 ≤ i ≤ r . Since the g i also commute and satisfy the
n
relations g i i = 1, the elements ϕ(x) and ϕ(y) can both be reduced to the product
Qr ki
i =1 g i , hence ϕ(x) = ϕ(y).
Clearly ϕ : x i 7→ g i for 1 ≤ i ≤ r , and trivially ϕ is a homomorphism. Also ϕ is
unique since any homomorphism sending x i 7→ g i for 1 ≤ i ≤ r must map according
to ϕ. This completes the proof.

60
Note this proof uses some of the same reasoning used in the proof that two groups
with the same presentations are isomorphic. Also, this proof captures the essence of
the proof of the Fundamental Theorem for Finitely Generated Abelian Groups.

E XERCISE 16. No finitely generated abelian group is divisible.

Proof. Let G = 〈g 1 , . . . , g r 〉 be a finitely generated abelian group. If each g i has finite

order, then G is finite and we are done (by Exercise 2.4.19, or directly by noting that
for k = |g i |, g k = 1 for all g ∈ G, hence no nontrivial element in G has a k-th root).
Q

If |g i | = ∞, then g i has no square root, hence G is not divisible.

Section 4

In the following exercises, for a group G, G 0 denotes the commutator subgroup of G.

E XERCISE 4,5. S 40 = A 4 and A 04 = V4 . For all n ≥ 5, S n0 = A 0n = A n .

Proof. Since S n is nonabelian but S n /A n = ∼ Z2 which is abelian, 1 < S 0 ≤ A n . It is

n
easily seen (by looking at the group lattice) that no nontrivial proper subgroup of A 4
is normal in S 4 , hence S 40 = A 4 . For n ≥ 5, A n is simple, so S n0 = A n . Note this shows
in particular that A n is characteristic in S n for all n ≥ 4.
Since A 4 is nonabelian but A 4 /V4 ∼ = Z3 is abelian, 1 < A 04 ≤ V4 . Now A 04 char
0
A 4 char S 4 , hence A 4 char S 4 . No nontrivial proper subgroup of V4 is characteristic
in S 4 , hence A 04 = V4 . For n ≥ 5, simplicity of A n forces A 0n = A n .

Note for 1 ≤ n ≤ 3, we also have S n0 = A n . Also for 1 ≤ n ≤ 2, A 0n = A n , but A 03 = 1.

Thus in particular we obtain: for all n ≥ 1, A n is characteristic in S n .

E XERCISE 7. Let P a nonabelian group with |P | = p 3 (p prime). Then P 0 = Z (P ).

Proof. Since P is a nontrivial p-group, Z (P ) 6= 1. Also p 2 does not divide |Z (P )|, lest
P /Z (P ) be cyclic, contradicting that P is nonabelian. Hence |Z (P )| = p.
It follows that |P /Z (P )| = p 2 , hence P /Z (P ) is abelian and P 0 ≤ Z (P ). Since P is
nonabelian, P 0 6= 1, thus we must have P 0 = Z (P ).

E XERCISE 10. A finite abelian group is the direct product of its Sylow subgroups.
α α
Proof. Let A be a finite abelian group. Set n = |A| and write n = p 1 1 · · · p k k (where
the p i are pairwise distinct primes). Let A i be the Sylow subgroup of A with |A i | =
α
p i i . Then A i E A, and by Lagrange’s Theorem, A i ∩ A j = 1 for i 6= j . Thus we have

A = A1 · · · Ak ∼
Y
= Ai

61
E XERCISE 11. Let G be a group and suppose G = H K for characteristic subgroups
H , K with H ∩ K = 1. Then Aut(G) ∼
= Aut(H ) × Aut(K ).

Proof. Note any automorphism of H or K may be extended to an automorphism

of G in a natural way. In this way, identify Aut(H ) and Aut(K ) with their naturally
corresponding subgroups of Aut(G).
Claim Aut(H ) E Aut(G). Indeed, for σ ∈ Aut(G) and τ ∈ Aut(H ), στσ−1 ∈ Aut(G),
and since H char G, στσ−1 maps H to itself and is the identity on G − H , so στσ−1 ∈
Aut(H ). By similar reasoning, Aut(K ) E Aut(G). Also Aut(H ) ∩ Aut(K ) = 1 because
G = H K and H ∩ K = 1. Indeed, if σ ∈ Aut(H ) ∩ Aut(K ), then σ is the identity on H
(since σ ∈ Aut(K )) and σ is the identity on K (since σ ∈ Aut(H )), hence σ = 1.
Finally Aut(G) = Aut(H )Aut(K ) since for σ ∈ Aut(G), σ = (σ|H ) ◦ (σ|K ) (this also
relies on G = H K and H ∩ K = 1). Hence we have

Aut(G) = Aut(H )Aut(K ) ∼

= Aut(H ) × Aut(K )

This result shows that if a group can be factored into disjoint subgroups, each of
which is preserved under automorphisms of the group, then any automorphism of
the group can be naturally factored into automorphisms of those subgroups.

E XERCISE 17. Let K be a cyclic normal subgroup of G. Then G 0 ≤ CG (K ).

Proof. Since K E G, NG (K ) = G. Thus G/CG (K ) is isomorphic to a subgroup of Aut(K ),

which is abelian since K is cyclic. Thus G 0 ≤ CG (K ).
Q
E XERCISE 18. Let K 1 , . . . , K n be nonabelian simple groups and set G = K i . If N E G,
then N = G I for some I ⊆ {1, . . . , n}.

Proof. Let N E G and set Ni = N ∩K i . Then Ni E K i , and so since K i is simple, Ni = 1

or Ni = K i . Note Z (K i ) = 1 since Z (K i ) E K i but Z (K i ) 6= K i .
Q
Claim N = Ni . Indeed, suppose x = (a 1 , . . . , a n ) ∈ N . If a i = 1, trivially a i ∈ Ni .
If a i 6= 1, then (because a i 6∈ Z (K i )) there exists y ∈ K i with [a i , y] 6= 1. It follows that
Q Q
1 6= [x, y] ∈ Ni , so Ni = K i and K i ≤ N . Thus x ∈ Ni . Since x was arbitrary, N ≤ Ni .
The reverse inclusion is trivial.
It follows that G = G I where I = { i | 1 ≤ i ≤ n and Ni 6= 1 }.

Section 5

For the following exercises, unless otherwise noted, H and K denote arbitrary groups,
ϕ : K → Aut(H ), and G = H oϕ K (with H and K treated as subgroups of G).

62
E XERCISE 1. C K (H ) = ker ϕ

Proof. From definitions and Theorem 10 we have

ker ϕ = { k ∈ K | ϕ(k) = 1 }
= { k ∈ K | (∀h ∈ H )[ϕ(k)(h) = h] }
= { k ∈ K | (∀h ∈ H )(khk −1 = h) }
= { k ∈ K | (∀h ∈ H )(kh = hk) }
= { k ∈ K | k ∈ CG (H ) }
= K ∩CG (H ) = C K (H )

E XERCISE 2. C H (K ) = N H (K )

Proof. Trivially C H (K ) ≤ N H (K ). Conversely suppose h ∈ N H (K ) and let k ∈ K be

arbitrary. Consider the commutator element [h, k] = h −1 k −1 hk. Since h ∈ N H (K ),
(h −1 k −1 h)k ∈ K , and since K acts on H by conjugation, h −1 (k −1 hk) ∈ H . Hence
[h, k] ∈ H ∩ K = 1, so [h, k] = 1 and h and k commute. Since k was arbitrary, h ∈
C H (K ). Since h was arbitrary, N H (K ) ≤ C H (K ), completing the proof.

E XERCISE 5. Let G = Hol(Z2 × Z2 ).

(a) G = H o K where H = Z2 × Z2 and K ∼

= S 3 . Thus |G| = 24.

Proof. Set H = Z2 ×Z2 and K = Aut(H ). Then G = H o K by definition, and so it

suffices to prove that K ∼ = S 3 . This follows from the observation that there are
exactly three elements of order 2 in H (if H = 〈x〉×〈x〉, these elements are (x, 1),
(1, x), and (x, x)), K acts on these elements, and the induced representation is
bijective. It follows that |G| = 4 · 3! = 24.

(b) G ∼
= S4.

Proof. Note |G : K | = 4, so the action of G on the coset space G/K induces a

representation ϕ : G → S 4 . We claim this representation is an isomorphism.
Since |G| = 24 = |S 4 |, it is sufficient to prove ϕ is faithful, i.e., ker ϕ = 1.
Note ker ϕ ≤ K . If k ∈ K , k 6= 1, then by definition of the holomorph there
exists some h ∈ H such that khk −1 = k ·h 6= h, that is, h and k do not commute
in G. But then we cannot have k(hK ) = hK , lest in particular h −1 kh ∈ K , so
(as in the proof of Exercise 2) [h, k] ∈ H ∩ K = 1, a contradiction. Thus k acts
nontrivially on G/K . Since k was arbitrary nontrivial, ker ϕ = 1 as desired.

63
E XERCISE 6 (I SOMORPHIC S EMIDIRECT P RODUCTS ). Suppose H is arbitrary, K is
cyclic, and ϕ1 and ϕ2 are homomorphisms from K to Aut(H ) (assumed injective
if K is infinite) such that ϕ1 (K ) is conjugate to ϕ2 (K ). Then H oϕ1 K ∼
= H oϕ2 K .

Proof. Suppose σϕ1 (K )σ−1 = ϕ2 (K ). Since K is cyclic, there exists some n such that
σϕ1 (k)σ−1 = ϕ2 (k)n for all k ∈ K .
Recall that for h ∈ H and k ∈ K , k acts on h in H oϕ1 K by conjugation with
khk −1 = ϕ1 (k)(h). From our intuitive understanding of conjugation, we know that
σϕ1 (k)σ−1 = ϕ2 (k)n = ϕ2 (k n ) acts ‘in the same way’ on σ(h), that is, k n acts ‘in the
same way’ on σ(h) in H oϕ2 K . This suggests the isomorphism

ψ : (h, k) 7→ (σ(h), k n )

Indeed, we claim ψ is an isomorphism. First note that ψ is a homomorphism since

if (h 1 , k 1 ) and (h 2 , k 2 ) are in H oϕ1 K , then

ψ (h 1 , k 1 )(h 2 , k 2 ) = ψ h 1 ϕ(k 1 )(h 2 ) , k 1 k 2

£ ¤ ¡ ¢

= σ(h 1 ϕ1 (k 1 )(h 2 )) , (k 1 k 2 )n
¡ ¢

= σ(h 1 )σϕ1 (k 1 )(h 2 ) , k 1n k 2n

¡ ¢

= σ(h 1 )σϕ1 (k 1 )σ−1 (σ(h 2 )) , k 1n k 2n

¡ ¢

= σ(h 1 )ϕ2 (k 1n )(σ(h 2 )) , k 1n k 2n

¡ ¢

= (σ(h 1 ), k 1n )(σ(h 2 ), k 2n )
= ψ(h 1 , k 1 )ψ(h 2 , k 2 )

We claim that ψ is bijective.

Indeed, if K is infinite, note that σ−1 ϕ2 (K )σ = ϕ1 (K ), hence there exists some m
such that σ−1 ϕ2 (k)σ = ϕ1 (k)m = ϕ1 (k m ) for all k ∈ K . This suggests that a two-sided
inverse for ψ is given by
ψ∗ : (h, k) 7→ (σ−1 (h), k m )
To prove that it is indeed an inverse, it is sufficient to prove k mn = k for all k ∈ K . We
obtain for all k ∈ K

ϕ1 (k mn ) = ϕ1 (k n )m = σ−1 ϕ2 (k n )σ = σ−1 (σϕ1 (k)σ−1 )σ = ϕ1 (k)

and similarly ϕ2 (k mn ) = ϕ2 (k). Thus since ϕ1 and ϕ2 are injective by hypothesis, we

must have k mn = k for all k ∈ K as desired.
If K is finite,2 note that since |ϕ1 (K )| = |ϕ2 (K )| we have

|K | |K |
|ker ϕ1 | = = = |ker ϕ2 |
|ϕ1 (K )| |ϕ2 (K )|

Therefore ker ϕ1 = ker ϕ2 since K is cyclic (and hence has a unique subgroup of this
order). Now set N = ker ϕi , K = K /N , and let ϕi : K → ϕi (K ) be the naturally induced
isomorphisms. Also let c σ denote conjugation by σ. Then ϕ = ϕ2 −1 c σ ϕ1 ∈ Aut(K ).
2 For this case I required some assistance from Professor Google.

64
But there is a natural surjective homomorphism Aut(K ) → Aut(K ), hence there is a
map k 7→ k r ∈ Aut(K ) whose image is ϕ. It follows from computation that c σ (ϕ1 (k)) =
ϕ2 (k)r for all k ∈ K , so n = r . Thus there exists m such that (k n )m = k for all k ∈ K ,
and hence the map ψ∗ above, with this m, is a two-sided inverse of ψ.
Thus in either case ψ is an isomorphism, completing the proof.

Note in particular if ϕ1 (K ) = ϕ2 (K ), then H oϕ1 K ∼

= H oϕ2 K . It follows from this that,
up to isomorphism, when constructing a semidirect product H o K with K cyclic, it
does not matter which generator of K is chosen to define the action on H , since in all
cases the representation of K will produce the same set of automorphisms. This is a
very useful observation when classifying groups.

E XERCISE 7 (G ROUPS OF ORDER 56). We classify groups G with |G| = 56.

(a) If G is abelian, then G is isomorphic to one of

Z 8 × Z 7 (∼
= Z56 ) Z2 × Z4 × Z7 Z2 × Z2 × Z2 × Z7

Proof. By the Fundamental Theorem. Note 56 = 23 · 7, hence the possible lists

of elementary divisors of G are {8, 7}, {2, 4, 7}, and {2, 2, 2, 7}, which correspond
to the direct products given above.
Since (8, 7) = 1, Z8 × Z7 ∼
= Z56 gives the cyclic case.

(b) G has a normal Sylow 2-subgroup or a normal Sylow 7-subgroup.

Proof. Recall n 7 ≡ 1(7) and n 7 |8, hence n 7 = 1 or n 7 = 8. If n 7 = 1, the Sylow

7-subgroup is normal. If n 7 = 8, then since distinct Sylow 7-subgroups have
trivial intersection, there are (7−1)·8 = 48 elements of order 7 in G. This leaves
56 − 48 = 8 elements remaining, namely the elements in the single, and hence
normal, Sylow 2-subgroup.

(c) We classify nonabelian G with a normal Sylow 7-subgroup.

Note that G =∼ H oϕ K where H ∼ = Z7 (and hence Aut(H ) ∼= Z6 ), |K | = 8, and
ϕ : K → Aut(H ). From the classification of groups of order 8 we know K is
isomorphic to one of

Z8 Z2 × Z4 Z2 × Z2 × Z2 D 8 Q8

We consider these cases in turn and, for each such K , find all possible maps ϕ
(up to a possible choice of generators in K , which does not change the iso-
morphism type by Exercise 6). In the first three cases we do not consider case
ϕ = 1 explicitly, since the trivial map gives rise to abelian groups which we
have already classified above. Note since (8, 6) = 2, if ϕ 6= 1, then ϕ(K ) must be
the unique subgroup of Aut(H ) of order 2.
If K ∼
= Z8 , then we must have ker ϕ = Z4 .

65
 If K ∼
= Z2 × Z4 , then we must have either ker ϕ = Z4 or ker ϕ = Z2 × Z2 .
If K ∼
= Z2 × Z2 × Z2 , then we must have ker ϕ ∼
= Z2 × Z2 .
If K ∼
= D 8 , then ker ϕ must be one of D 8 , Z4 , or Z2 × Z2 .
If K ∼
= Q 8 , then ker ϕ must be one of Q 8 or Z4 (recall there are no subgroups
of Q 8 isomorphic to Z2 × Z2 ).
In addition, each one of these cases is realized, and gives rise to a distinct
semidirect product type (since for each distinct type of K , the kernels listed
are of different types).
Thus there are nine possible isomorphism types for nonabelian G with normal
Sylow 7-subgroup.
(d),(e) We classify (nonabelian) G with a nonnormal Sylow 7-subgroup.
We know that G must have a normal Sylow 2-subgroup, so G ∼= H oϕ K where
|H | = 8, K ∼
= Z7 , and ϕ : K → Aut(H ).
We claim H ∼ = Z2 ×Z2 ×Z2 . Indeed, if k ∈ Z7 , k 6= 1, then k acts nontrivially on H
by conjugation. By the Orbit-Stabilizer Theorem then, the orbit of any element
in H moved by k must have order 7. Hence the 7 nonidentity elements in H
must all be in a single orbit, and so have the same order. The only possibility
is the one stated by the claim.
We know Aut(H ) ∼ = GL 3 (F2 ), so |Aut(H )| = 168 = 23 ·3·7. Thus in particular there
exists an element in Aut(H ) of order 7, so a semidirect product of the current
form does exist. In addition, all subgroups of order 7 in Aut(H ) are conjugate
by Sylow’s Theorem, so by Exercise 6, this gives only one isomorphism type.

It follows from this exercise that there are exactly thirteen distinct isomorphism
types for groups of order 56.

E XERCISE 8 (G ROUPS OF ORDER 75).

Suppose G is a group with |G| = 75 = 3 · 52 . Recall from Sylow’s Theorem that
n 5 ≡ 1(5) and n 5 |3, hence n 5 = 1 and the Sylow 5-subgroup of G is normal. It follows
that G ∼= H oϕ K where |H | = 52 , K ∼ = Z3 , and ϕ : K → Aut(H ).
Now H is abelian since it has order p 2 (p = 5), so H is isomorphic to one of Z25
or Z5 × Z5 . If ϕ = 1, G is isomorphic to one of the abelian groups

Z3 × Z25 (∼
= Z75 ) Z3 × Z5 × Z5

Suppose ϕ 6= 1, so Aut(H ) has a subgroup of order 3. Recall |Aut(Z25 )| = 5(5 − 1) =

22 · 5, so we cannot have H ∼ = Z25 in this case. If H ∼
= Z5 × Z5 = E 52 , then Aut(H ) ∼
=
2 2 5
GL 2 (F5 ), so |Aut(H )| = (5 −1)(5 −5) = 2 ·3·5. In particular, Aut(H ) has a subgroup of
order 3, and every subgroup of order 3 in Aut(H ) is conjugate (since they are Sylow
3-subgroups). Thus this case is possible, and (by Exercise 6) gives rise to a single
nonabelian isomorphism type for G.

66
E XERCISE 12 (G ROUPS OF ORDER 20).
Suppose G is a group with |G| = 20 = 22 · 5. By Sylow’s Theorem, n 5 ≡ 1(5) and
n 5 |4, hence n 5 = 1 and G ∼= H oϕ K where H ∼ = Z5 , |K | = 4 (so K ∼
= Z4 or K ∼ = Z2 × Z2 ),
∼
and ϕ : K → Aut(H ) = Z4 .
If ϕ = 1, G is one of the abelian groups Z5 × Z4 (∼ = Z20 ) or Z5 × Z2 × Z2 .
If ϕ 6= 1, then if K ∼
= Z4 , either ker ϕ = 1 and G ∼ = Hol(Z5 ) or else ker ϕ ∼ = Z2 . If
∼ ∼
K = Z2 × Z2 , then we must have ker ϕ = Z2 .
Since all of these cases are possible and nonisomorphic, there are exactly five
isomorphism types for groups of order 20.

E XERCISE 15 (G ROUPS OF ORDER 2p 2 ).

Suppose G is a group with |G| = 2p 2 , p prime. If p = 2, then |G| = 23 = 8, an order
which we have already classified, so assume p 6= 2. From Sylow’s Theorem, n p ≡ 1(p)
and n p |2, hence n p = 1 and G ∼ = H oϕ K where |H | = p 2 , K ∼ = Z2 , and ϕ : K → Aut(H ).
Since H is abelian (order p 2 ), H is isomorphic to one of Z p 2 or Z p × Z p . If ϕ = 1,
G is isomorphic to one of Z p 2 × Z2 (∼ = Z2p 2 ) or Z p × Z p × Z2 .
Suppose ϕ 6= 1. If H ∼ = Z p 2 , then Aut(H ) ∼ = Z p(p−1) , and 2 must divide p − 1. Since
Z p(p−1) is cyclic, it has a unique subgroup of order 2, hence this case gives rise to a
single isomorphism type. Note this case is possible (take p = 3).
If H ∼= Z p × Z p = E p 2 , then Aut(H ) ∼ = GL 2 (Fp ). We claim that in GL 2 (Fp ), every
element of order 2 is conjugate (similar) to one of
· ¸ · ¸
1 −1
−1 −1

Note that this is equivalent to proving that for any A ∈ GL 2 (Fp ) with |A| = 2, there is
a change of basis under which A becomes one of the above. (Note that since p 6= 2,
and hence 2 6= 0, the two matrices above are not conjugate, since the one at right has
no eigenvectors but the one at left does.)
Suppose A ∈ GL 2 (Fp ) with |A| = 2. Note3 that for any v ∈ Z p2 ,

A(v + Av) = Av + A 2 v = Av + v = v + Av
A(v − Av) = Av − A 2 v = Av − v = −(v − Av)

Let (v, w) be a basis for the space. Note it cannot be true that both v + Av and v − Av
are in the span of w, lest
1
v = ((v + Av) + (v − Av))
2
is also in the span of w, contradicting the linear independence of (v, w). (Note the
above equation relies on the fact that p 6= 2, so 2 6= 0 and 1/2 is defined.) Let v ∗ be
whichever is not in the span of w. Then (v ∗ , w) is a basis and Av ∗ = ±v ∗ . Similarly
there is w ∗ with Aw ∗ = ±w ∗ such that (v ∗ , w ∗ ) is a basis of the space. This proves
the conjugacy claim above.
It follows that there are at most two (and at least one, taking say p = 3) possible
isomorphism types for G in this case.

3 For the following observations I required some assistance from Professor Google.

67
E XERCISE 18. Let H be any group. There exists a group G containing H as a normal
subgroup such that any automorphism of H is witnessed as an inner automorphism
in G (restricted to H ).

Proof. Take G = Hol(H ) = H o Aut(H ).

E XERCISE 19 (H OLOMORPHS AS N ORMALIZERS ). Let H be a finite group with |H | = n,

K = Aut(H ), and G = Hol(H ) = H o K . Let G act on the coset space G/K by left
multiplication with representation π : G → S n (since |G : K | = n).

(a) The elements of H are coset representatives in G/K , and if we identify H and G/K ,
π|H is just the left regular representation of H .

Proof. Let h 1 , h 2 ∈ H . If h 1 K = h 2 K , then h 2−1 h 1 K = K , so h 2−1 h 1 ∈ H ∩ K = 1,

and hence h 1 = h 2 . Thus we may identify H and G/K by the natural projection
h 7→ hK . Trivially then π|H is just the left regular representation of H .

(b) NS n (π(H )) = π(G). Therefore in general the normalizer of the (left) regular
representation of a group H is isomorphic to Hol(H ).

Proof. Note π(H ) E π(G) since H E G, hence π(G) ≤ N (π(H )). Because H is
finite, it is sufficient to prove that |π(G)| = |N (π(H ))|.
We need |N (π(H ))| ≤ |π(G)|.
Recall N (π(H ))/C (π(H )) is isomorphic to a subgroup of Aut(π(H )). Since π is
injective, Aut(π(H )) ∼
= Aut(H ) = K . By Exercise 4.3.36, C (π(H )) ∼
= H . Hence

|N (π(H ))|
divides |K |
|H |

Thus |N (π(H ))| divides |H ||K | = |G|, so |N (π(H ))| ≤ |π(G)| as desired.
Thus N (π(H )) = π(G) ∼= G = Hol(H ). The general result follows from (a).

(c) The normalizer of a group generated by an n-cycle in S n is isomorphic to Hol(Zn )

and has order nϕ(n).

Proof. If σ is the n-cycle and Zn = 〈x〉, the map x 7→ σ realizes 〈σ〉 as a (left)
regular representation of Zn under some numbering of Zn . Thus by part (b),
NS n (σ) ∼
= Hol(Zn ) = Zn oAut(Zn ), and since |Aut(Zn )| = ϕ(n), |NS n (σ)| = nϕ(n).

Note this exercise shows how holomorphs arise naturally from permutation groups.
Also, this and Exercise 4.3.36 together give us characterizations of the centralizers
and normalizers of regular permutation representations.

68
E XERCISE 23 ( W REATH P RODUCT ). Let K and L be groups with ρ : K → S n for some
positive integer n.
Let H = n L and ϕ : K → Aut(H ) be naturally induced from ρ by letting the
Q

elements of K permute the factors of H (see Exercise 1.8). Then the wreath product
of L by K is defined by
L o K = H oϕ K
If ρ is not mentioned, it is assumed to be the left regular representation of (finite) K .
(a) Suppose K and L are finite. Then |L o K | = |L||K | |K |. Proof: Immediate.
(b) Let p be a prime. Then Z p o Z p is a nonabelian group of order p p+1 isomorphic
to a subgroup of S p 2 .

Proof. The group Z p o Z p is nonabelian since the left regular representation

of Z p is nontrivial, and has order p p+1 by (a).
The p copies of Z p in the direct product can be seen as producing a grid
of p 2 elements, with the elements of each factor arranged in vertical columns.
The wreath product Z p o Z p acts naturally on this grid as follows: the element
α αp αp+1
(x 1 1 , · · · , x p , x p+1 ) cyclically permutes the elements in column i by αi for
1 ≤ i ≤ p, and then cyclically permutes the columns themselves by αp+1 . We
thus obtain a faithful representation Z p o Z p → S p 2 , so Z p o Z p is isomorphic to
a subgroup of S p 2 .

Note examples like (b) might shed some light on the name ‘wreath product’. In that
example, each of the factors in the direct product can be seen as a little circle, and
the factors themselves are cyclically permuted in a larger circle.

Chapter 6
Section 1

E XERCISE 1. Let G be a group. Then Zi (G) char G for all i .

Proof. We proceed by induction on i .

Case i = 0: Z0 (G) = 1 char G.
Case i = 1: Z1 (G) = Z (G), and we claim Z (G) char G. Indeed, fix σ ∈ Aut(G) and
suppose x ∈ Z (G) and g ∈ G. Let σ(h) = g . Then

g σ(x) = σ(h)σ(x) = σ(hx) = σ(xh) = σ(x)σ(h) = σ(x)g

Letting g vary we see σ(x) ∈ Z (G), then letting x vary we see σ[Z (G)] ⊆ Z (G). Now
substituting σ−1 for σ, we have σ−1 [Z (G)] ⊆ Z (G), so Z (G) = σ[σ−1 [Z (G)]] ⊆ σ[Z (G)],
so σ[Z (G)] = Z (G). Since σ was arbitrary, Z (G) char G as claimed.
Now if Zi (G) char G for some i ≥ 1, set G = G/Zi (G). If σ ∈ Aut(G), σ naturally
induces a map

σ:G →G
g 7→ σ(g )

69
This map is well-defined since if x = y, that is, x Zi (G) = y Zi (G), then

σ(x) = σ(x)Zi (G) = σ[x Zi (G)] = σ[y Zi (G)] = σ(y)Zi (G) = σ(y)

where the second and fourth equalities hold since Zi (G) char G. It is immediate that
σ is a homomorphism, and σ−1 is a two-sided inverse of σ, hence σ ∈ Aut(G).
Now it follows from the reasoning above that σ[Z (G)] = Z (G). We claim that
σ[Zi +1 (G)] = Zi +1 (G). Indeed, recall by definition Zi +1 (G) = Z (G). Thus if g ∈ Zi +1 ,
then g ∈ Z (G), hence σ(g ) = σ(g ) ∈ Z (G), that is, σ(g ) ∈ Zi +1 . Since g was arbitrary,
σ[Zi +1 (G)] ⊆ Zi +1 (G), and as above it follows that σ[Zi +1 (G)] = Zi +1 (G) as claimed.
Thus Zi +1 char G. By induction, the result is true for all i .

E XERCISE 2,4,5 (N ILPOTENT G ROUPS ARE L IKE p -G ROUPS ). Let G be nilpotent.

(a) If 1 < H E G, then H ∩ Z (G) 6= 1. In particular, if G 6= 1, then Z (G) 6= 1.4

Proof. If H ≤ Z (G), the result holds, so assume that H 6≤ Z (G). Then we have
Z (G) < H Z (G) E G, so by the Lattice Theorem 1 < H E G where G = G/Z (G).
By induction on nilpotence class, we may assume H ∩ Z (G) 6= 1. Now it follows
from the Lattice Theorem again that H Z (G) ∩ Z2 (G) 6≤ Z (G), which implies
H ∩ Z2 (G) 6= 1. Since H E G, [G, H ] ≤ H (Proposition 5.7(2)), and by definition
of Z2 (G), [G, Z2 (G)] ≤ Z (G). Therefore we have [G, H ∩ Z2 (G)] ≤ H ∩ Z (G). If
[G, H ∩ Z2 (G)] 6= 1, the result holds, otherwise that means H ∩ Z2 (G) ≤ Z (G), so
again H ∩ Z (G) 6= 1 and the result holds.

(b) If H < G, then H < NG (H ).

Proof. Suppose H < G. If Z (G) 6≤ H , then H < 〈H , Z (G)〉 ≤ NG (H ). If Z (G) ≤ H ,

set G = G/Z (G). By the Lattice Theorem, H < G. Since Z (G) 6= 1, by induction
on nilpotence class we may assume H < NG (H ). Then since NG (H ) = NG (H ),
H < NG (H ) by the Lattice Theorem.

(c) If G is finite and M ≤ G is maximal, then |G : M | = p for some prime p.

Proof. Suppose G is finite and M is a maximal subgroup of G. Then M E G

by Proposition 7, hence G = G/M is defined and |G : M | = |G|. Note |G| > 1
since M < G. If |G| is composite, then (by Cauchy’s Theorem, say) there exists
a subgroup 1 < H < G, so M < H < G by the Lattice Theorem, contradicting
the maximality of M . Therefore |G| is prime, as desired.

Together with the fact that all p-groups are nilpotent (Proposition 2), these results
exhibit strong connections between p-groups and nilpotent groups in general. For
this reason it is often convenient to think about (simpler) p-groups when working
with nilpotent groups.
4 For this problem, I required some assistance from Professor Google.

70
E XERCISE 3. Let G be a finite group. Then G is nilpotent iff G has a normal subgroup
of every order dividing |G|, and G is cyclic iff G has a unique subgroup of every order
dividing |G|.

Proof. If G has a normal subgroup of every order dividing |G|, then in particular
G has a normal (and hence unique) Sylow p-subgroup for all primes p dividing |G|.
Thus all Sylow p-subgroups of G are normal, so G is nilpotent by Theorem 3.
α α
Conversely, if G is nilpotent, write |G| = p 1 1 · · · p k k (p i prime), so by Theorem 3
β β
G = P 1 × · · · × P k where P i ∈ Sylp i (G). If n divides |G|, write n = p 1 1 · · · p k k . Then by
β
Theorem 1(3), we know each P i has a normal subgroup Ni with |Ni | = p i i . Now set
N = N1 × · · · × Nk . Then N E G and |N | = n, as desired.
If G is cyclic, we know G has a unique subgroup of each order dividing |G| by
Theorem 2.7(3). To prove the converse, we exhibit for any noncyclic G two distinct
subgroups having the same order. So consider an arbitrary noncyclic group G. If G is
abelian, then by the fundamental theorem and Proposition 5.6(2), we must be able
to write G as a direct product of cyclic groups where, for some prime p, at least two
of the factors are nontrivial p-groups. From these p-groups we can select elements
of order p which generate distinct subgroups of order p.
If G is nilpotent but not abelian, proceed by induction. Note G/Z (G) is also
nilpotent but not cyclic, lest G be abelian. Hence we may assume G/Z (G) contains
distinct subgroups of the same order. The complete preimages of these subgroups
in G are distinct subgroups of G of the same order.
Finally if G is not nilpotent, then by Theorem 3 at least one of its Sylow subgroups
is nonnormal and hence not unique. Therefore in all cases, G has distinct subgroups
of the same order.

E XERCISE 6. G is nilpotent iff G/Z (G) is nilpotent.

Proof. Set G = G/Z (G). We first establish a natural relationship between the upper
central series in G and the upper central series in G:

Zi +1 (G) = Zi (G) (i ≥ 0)

To prove this, we proceed by induction on i . The result is immediate for cases i = 0, 1.

If the result holds for some i ≥ 1, we claim that the result also holds for i + 1, that is,
Zi +2 (G) = Zi +1 (G). Note both Zi +2 (G) and Zi +1 (G) are subgroups of G. Our idea is
to choose N E G with N ≤ Zi +2 (G) ∩ Zi +1 (G) such that Zi +2 (G)/N = Zi +1 (G)/N , so
that our claim then follows from the Lattice Theorem.
Choose N = Zi (G). Then N E G and N ≤ Zi +1 (G). By hypothesis, N = Zi +1 (G),
and since Zi +1 (G) ≤ Zi +2 (G), we have N ≤ Zi +2 (G) by the Lattice Theorem. Now

Zi +1 (G)/N = Zi +1 (G)/Zi (G)

= Z (G/Zi (G)) by definition of Zi +1 (G)
= Z (G/Zi +1 (G))

On the other hand, Zi +2 (G)/N = Zi +2 (G)/Zi +1 (G). Therefore we must prove that
Zi +2 (G)/Zi +1 (G) = Z (G/Zi +1 (G)).

71
 We know that Zi +2 (G)/Zi +1 (G) = Z (G/Zi +1 (G)) by definition of Zi +2 (G). And by
the Third Isomorphism Theorem, G/Zi +1 (G) ∼ = G/Zi +1 (G). Using this isomorphism,
and the fact that centers are preserved under isomorphism, it can be seen that the
required equality holds.
Therefore our claim holds for i + 1. By induction, our result holds for all i ≥ 0.
Returning to the problem, note for i ≥ 0 that Zi +1 (G) = G iff Zi (G) = G by our
result and the Lattice Theorem. Hence G is nilpotent iff G is nilpotent, as desired.

Note the proof gives us more information: for n > 0, G is of nilpotence class n + 1 iff
G/Z (G) is of nilpotence class n.

E XERCISE 7. Let G be nilpotent. If H ≤ G, then H is nilpotent. If H E G, then G/H is

also nilpotent.

Proof. We use the lower central series characterization of nilpotence.

Suppose H ≤ G. We claim H i ≤ G i for i ≥ 0. Indeed, H 0 = H ≤ G = G 0 . If i ≥ 0
and H i ≤ G i , then by definitions it follows that

H i +1 = [H , H i ] ≤ [G,G i ] = G i +1

Thus the claim holds for all i ≥ 0 by induction. Since G is nilpotent, G n = 1 for
some n, so H n = 1 and H is nilpotent.
Suppose now ϕ : G → K is a surjective homomorphism. We claim ϕ[G i ] = K i for
i ≥ 0. Note ϕ[G 0 ] = ϕ[G] = H = H 0 since ϕ is surjective. If i ≥ 0 and ϕ[G i ] = H i , then

ϕ[G i +1 ] = ϕ[[G,G i ]] = [ϕ[G], ϕ[G i ]] = [H , H i ] = H i +1

The second equality follows since if x ∈ G and y ∈ G i , then ϕ([x, y]) = [ϕ(x), ϕ(y)] by
the definition of the commutator. Hence the claim holds for all i ≥ 0 by induction.
Since G is nilpotent, G n = 1 for some n, hence K n = 1 and K is nilpotent. Thus we
see that the homomorphic image of a nilpotent group is nilpotent.
If H E G, then G/H is the homomorphic image of G under a natural projection,
so G/H is nilpotent.

It is not true in general that if H E G is nilpotent and G/H is nilpotent, then G is

nilpotent. Indeed, S 3 is not nilpotent, but 〈(123)〉 E S 3 is cyclic, hence abelian and
nilpotent, and S 3 /〈(123)〉 is cyclic of order 2, hence also nilpotent. This differentiates
nilpotent groups from solvable groups (Proposition 10(3)).

E XERCISE 8. Let p be prime and P a nonabelian group of order p 3 . Then |Z (P )| = p

and P /Z (P ) ∼
= Zp × Zp .

Proof. We know that |Z (P )| = p α for 0 ≤ α ≤ 3 by Lagrange’s Theorem, and α > 0 by

Theorem 1(1). If α > 1, then P /Z (P ) is cyclic, contradicting that P is nonabelian.
Therefore |Z (P )| = p, and |P /Z (P )| = p 2 . Since every group of order p 2 is abelian,
P /Z (P ) is abelian, and either P /Z (P ) ∼
= Z p 2 or P /Z (P ) ∼
= Z p × Z p (Corollary 4.9). The
first case is impossible since P is nonabelian, hence P /Z (P ) ∼ = Z p ×Z p as desired.

72
E XERCISE 9. Let G be finite. Then G is nilpotent iff for all x, y ∈ G if (|x|, |y|) = 1 then
x y = y x.

Proof. Suppose G is nilpotent. By Theorem 3(4), we may assume G = P 1 × · · · × P k

where P i ∈ Sylp i (G) for 1 ≤ i ≤ k. If x, y ∈ G with (|x|, |y|) = 1, write x = (x 1 , . . . , x k ) and
y = (y 1 , . . . , y k ), so |x| = lcm(|x 1 |, . . . , |x k |) and |y| = lcm(|y 1 |, . . . , |y k |). If x i 6= 1, then
y i = 1, lest p i is a common divisor of |x i | and |y i |, hence also of |x| and |y|. Thus
x i y i = y i x i for all 1 ≤ i ≤ k, so x y = y x as desired.
Conversely, suppose x y = y x for all x, y ∈ G with (|x|, |y|) = 1. Let H < G. Since
G is finite, |H | < |G|, and thus by Cauchy’s Theorem we may choose x ∈ G with |x| = p
for some prime p not dividing |H |. If y ∈ H , then y divides |H |, so (|x|, |y|) = 1, and
hence by assumption x y = y x. Since y was arbitrary, this shows x ∈ CG (H ) ≤ NG (H ),
and hence H < NG (H ). Since H was arbitrary, G is nilpotent by Theorem 3(2).

E XERCISE 10. D 2n is nilpotent iff n is a power of 2.

Proof. Suppose n is a power of 2 and write n = 2k . Since D 2n is trivially nilpotent

(k−1)
for k ≤ 1, we may assume k > 1. By Exercise 2.2.7, Z (D 2n ) = {1, r 2 }, hence by
Exercise 3.1.34, D 2n /Z (D 2n ) ∼ D
= 2·2k−1 . Now by induction we may assume D 2·2k−1 is
nilpotent, hence D 2n is nilpotent by Exercise 6.
Conversely, suppose n is not a power of 2. Then p|n for some prime p 6= 2, and
x = r (n/p) is an element in D 2n of order p. Now x is not in Z (D 2n ), hence x cannot
commute with s (since x commutes with r ), but (|x|, |s|) = (p, 2) = 1. Therefore D 2n is
not nilpotent by Exercise 9.

E XERCISE 12. We find the upper and lower central series for A 4 and S 4 .

Proof. By definition Z0 (A 4 ) = 1, and by examining the lattice of A 4 (Exercise 3.5.8)

and checking conjugates (Proposition 4.10), we see Z1 (A 4 ) = 1. Hence Zn (A 4 ) = 1
for all n ≥ 0. Similarly, Z0 (S 4 ) = 1, and by constructing conjugates we see Z1 (S 4 ) = 1,
hence Zn (S 4 ) = 1 for all n ≥ 0 as well. This establishes the two upper central series.
To find the lower central series of A 4 , note by definition A 04 = A 4 . Recall from
the lattice and Exercise 3.5.9 that V4 E A 4 . Now A 4 /V4 ∼ = Z3 is abelian, hence by
definition and Proposition 5.7(4), A 14 = [A 4 , A 4 ] ≤ V4 . Since A 14 is characteristic in A 4
(Exercise 14), A 14 = V4 or A 14 = 1, but since A 4 is not abelian, we must have A 14 = V4 .
Now A 24 = [A 4 ,V4 ], so again the only possibilities are A 24 = V4 or A 24 = 1. But it is
easy to check that not everything in V4 commutes with everything in A 4 (that is,
V4 6≤ Z (A 4 ) = 1), hence A 24 = V4 . It follows that A n4 = V4 for all n ≥ 1.
Similarly S 40 = S 4 , and since A 4 E S 4 and S 4 /A 4 ∼ = Z2 is abelian, we must have
1
S 4 = [S 4 , S 4 ] ≤ A 4 . Now V4 = [A 4 , A 4 ] ≤ [S 4 , S 4 ], but it is easy to check that V4 E
6 S4,
hence S 41 = A 4 . We then have S 42 = [S 4 , A 4 ], so again V4 ≤ S 42 ≤ A 4 , so again we must
have S 42 = A 4 . It follows that S 4n = A 4 for all n ≥ 1.
In particular, these results show that neither A 4 nor S 4 is nilpotent.

Note by this exercise and Exercise 3.5.10, A 4 is solvable but not nilpotent.

73
E XERCISE 13. We find the upper and lower central series for A n and S n , for n ≥ 5.

Proof. Fix n ≥ 5. Recall A n is simple (Theorem 4.24), that is, A n has no nontrivial
proper normal subgroups. We know that Z0 (A n ) = 1. By Exercise 14, we must have
Z1 (A n ) = 1 or Z1 (A n ) = A n . But A n is not abelian, hence Z1 (A n ) = 1 and it follows
that Zk (A n ) = 1 for all k ≥ 0. It is trivial that Zk (S n ) = 1 for all k ≥ 0.
We know A 0n = A n , and since A n is simple, A 1n = A n or A 1n = 1. Again since A n is
not abelian, we must have A 1n = A n , and it follows that A kn = A n for all k ≥ 0. Now
S n0 = S n , but A n E S n and S n /A n ∼ = Z2 is abelian, hence S n1 ≤ A n . Since S n is not
abelian, the simplicity of A n forces S n1 = A n . Now S n2 = [S n , A n ] ≤ A n . Again by the
simplicity of A n , and the fact that A n 6≤ Z (S n ) = 1, we must have S n2 = A n . It follows
that S nk = A n for all n ≥ 1.
In particular, these results show that neither A n nor S n is nilpotent.

E XERCISE 14,17. Let G be a group. For all i ≥ 0, G i char G and G (i ) char G.

Proof. By induction. Trivially G 0 = G (0) = G char G. If G i char G and G (i ) char G, let

σ ∈ Aut(G). Then

σ[G i +1 ] = σ[[G,G i ]] by definition of G i +1

= [σ[G], σ[G i ]] by definition of commutators
i
= [G,G ] by induction
i +1
=G by definition of G i +1

Similarly σ[G (i +1) ] = G (i +1) . Since σ was arbitrary, this shows that G i +1 char G and
G (i +1) char G. By induction, the result holds for all i ≥ 0.

E XERCISE 16. Q has no maximal subgroups.

Proof. Suppose towards a contradiction that H < Q is maximal. Fix q ∈ Q − H . Then

by maximality of H , we have H < 〈H , q〉 = Q. If r ∈ Q is arbitrary, then r = αq + h for
some α ∈ Z and h ∈ H .
Write q/2 = αq + h for α ∈ Z and h ∈ H . Then (1 − 2α)q = 2h ∈ H , and hence
|1 − 2α|q ∈ H (since H is a subgroup). Note 1 − 2α 6= 0 since α ∈ Z, so |1 − 2α| > 0
and there exists a positive integral multiple of q in H . Let µ ∈ Z be least positive with
µq ∈ H . By choice of q, µ > 1. We claim that the cosets

H , q + H , . . . , (µ − 1)q + H

partition Q. Indeed, if r ∈ Q, write r = βq +h 0 . By division with remainder, β = ρ +δµ

for some δ, ρ ∈ Z with 0 ≤ ρ < µ. Hence

r = βq + h 0 = (ρ + δµ)q + h 0 = ρq + (δµq + h 0 ) = ρq + h 00

where h 00 = δµq + h 0 ∈ H . Thus r + H = ρq + (h 00 + H ) = ρq + H , so since 0 ≤ ρ ≤ µ − 1,

r + H is one of the cosets above. Since r was arbitrary, this establishes the claim.
But we know Q has no proper subgroups of finite index (Exercise 3.2.21), so this
is a contradiction. Thus Q has no maximal subgroups.

74
E XERCISE 20. Let G be finite, P ∈ Sylp (G), and N E G with (|N |, p) = 1.

(a) NG (P ) = NG (P )

Proof. Note N < N P and N P = P , so by the Lattice Theorem

NG (N P ) = NG (N P ) = NG (P )

We claim N NG (P ) = NG (N P ), so that NG (P ) = N NG (P ) = NG (N P ) = NG (P ),
establishing the desired result.
Indeed, note that N P E N (N P ) and P ∈ Sylp (N P ) since p does not divide |N |.
Therefore by Frattini’s Argument, N NG (P ) = (N P )NG (P ) = NG (N P ), which is
precisely the claim.

Note the lattice of G naturally suggests N NG (P ) = NG (N P ): this just means the

chain N ≤ N P ≤ NG (N P ) is the lifting by N of the chain 1 ≤ P ≤ NG (P ). Also,
this proof shows the utility of Frattini’s Argument: knowing P ≤ N P E NG (N P )
with P a Sylow subgroup of N P , we use the fact that P is “somewhat normal”
in NG (N P ) to lift NG (P ) up to NG (N P ).

For a group G, the Frattini subgroup Φ(G) of G is defined to be the intersection of all
maximal subgroups of G (if G has no maximal subgroups, Φ(G) = G). The following
exercises establish some basic properties of the Frattini subgroup.

E XERCISE 21. Let G be a group. Then Φ(G) char G.

Proof. If σ ∈ Aut(G) and M is maximal in G, then σ[M ] is maximal in G. Therefore

σ[Φ(G)] = σ[ M ] = σ[M ] = Φ(G)

\ \
M M

where M varies over maximal subgroups of G. Since σ was arbitrary, Φ(G) char G.

E XERCISE 24. Let G be a group. Call an element x ∈ G a nongenerator if for all H < G,
〈H , x〉 < G. Then Φ(G) = { x ∈ G | x is a nongenerator }.

Proof. If x 6∈ Φ(G), then there exists some maximal M < G such that x 6∈ M . But then
M < 〈M , x〉 = G by maximality of M , hence x is not a nongenerator.
Conversely, if x is not a nongenerator, there exists H < G with 〈H , x〉 = G. Now
if H ≤ M < G with M maximal, then x 6∈ M lest G = 〈H , x〉 ≤ M < G, a contradiction.
Therefore we assume H is not contained in a maximal subgroup, so for all K with
H ≤ K < G, we may choose C (K ) with K < C (K ) < G. Set C (G) = G. Now define
recursively on the ordinals

H0 = H
Hα+ = C (Hα )
[
Hλ = Hα (λ a limit ordinal)
α<λ

75
By induction, this forms an ascending chain of subgroups in G, and there must exist
a limit ordinal µ with G = α<µ Hα and H ≤ Hα < G for all α < µ. Now x ∈ α<µ Hα ,
S S

hence x ∈ Hα for some α < µ. But then G = 〈H , x〉 ≤ Hα < G, a contradiction. Hence

this case is impossible, so x 6∈ Φ(G) by the previous case.

E XERCISE 25. Let G be a finite group. Then Φ(G) is nilpotent.

Proof. Let P ∈ Sylp (Φ(G)). We claim that P E Φ(G). Indeed, by Frattini’s argument
(Proposition 6), G = Φ(G)NG (P ). Now if NG (P ) < G, then NG (P ) ≤ M < G for some
maximal subgroup M . But then since Φ(G) ≤ M , we have G = Φ(G)NG (P ) ≤ M < G—
a contradiction. Therefore P E G, so in particular P E Φ(G) as claimed. Since P was
arbitrary, Φ(G) is nilpotent by Theorem 3(3).

E XERCISE 26 (B URNSIDE ’ S B ASIS T HEOREM ). Let p be prime, P be a finite p-group,

and P = P /Φ(P ).5

(a) P is elementary abelian.

Proof. If M is maximal in P , then M E P and |P : M | = p by Theorem 1(5),

hence P /M is cyclic of order p. It follows that P 0 ≤ M (by Proposition 5.7(4))
and x p ∈ M (that is, x p = 1) for all x ∈ P . Since M was arbitrary, P 0 ≤ Φ(P ) and
x p ∈ Φ(P ) for all x ∈ P , hence P /Φ(P ) is elementary abelian.

(b) If N E P and P /N is elementary abelian, then Φ(P ) ≤ N .

Proof. Write Pe = P /N . Note Φ(Pe) = 1 since Pe is elementary abelian (indeed,

Q
write Pe = Z p , and obtain maximal subgroups by dropping a Z p factor from
each position). Now

Φ(P f = Φ(Pe) = e
\ \
‚) = ‚ M≤ M 1
M M

where M varies over maximal subgroups of P . Hence Φ(P ) ≤ N .

(c) Burnside’s Basis Theorem: The set {y 1 , . . . , y k } is a minimal generating set for P
iff {y 1 , . . . , y k } is a basis of P (as a vector space over Fp ).

Proof. First we claim {y 1 , . . . , y k } generates P iff {y 1 , . . . , y k } spans P . Indeed,

the forward direction is immediate. Conversely, if {y 1 , . . . , y k } spans P , then
by the Lattice Theorem, P = 〈y 1 , . . . , y k 〉Φ(P ). But since Φ(P ) consists of non-
generators (Exercise 24), this implies P = 〈y 1 , . . . , y k 〉 as claimed.
It follows that the size of a minimal generating set is the same for both P and P .
Since a minimal generating set in P is just a basis of P (with size equal to the
dimension of P ), the desired result follows.

(d) If P is cyclic, P is cyclic. Moreover, if P /P 0 is cyclic, P is cyclic.

5 For this exercise, I required some assistance from Professor Google.

76
 Proof. If P is cyclic, then P is spanned by a single element, so by Burnside’s
Basis Theorem, P is generated by a single element, that is, P is cyclic.
If P /P 0 is cyclic, then P 0 = Φ(P ) by parts (a) and (b), hence P is cyclic, and P is
cyclic by the previous case.

For a group G, a minimal normal subgroup of G is a subgroup M with 1 < M E G

such that there do not exist any subgroups N with 1 < N < M and N E G.

E XERCISE 31. Let G be a finite solvable group and suppose M is a minimal normal
subgroup of G. Then M is an elementary abelian p-group for some prime p.

Proof. By Proposition 10(1), M is solvable, hence M 0 < M . But M 0 char M E G,

hence M 0 E G. By minimality of M , this means M 0 = 1, so M is abelian.
Let p divide |M | and set M p = { x ∈ M | x p = 1 }. Then M p is a subgroup of M since
M is abelian (Exercise 5.2.8), and it is immediate that M p char M , hence M p E G.
Since M p 6= 1 (by Cauchy’s Theorem, say), M p = M , hence M is elementary abelian.

E XERCISE 32. Let G be a finite solvable group and M be a maximal subgroup of G.

Then M has prime power index.

Proof. Let N be a minimal normal subgroup of G and set G = G/N .

If N ≤ M , then by the Lattice Theorem, M is maximal in G and |G : M | = |G : M | is
a prime power by induction. If N 6≤ M , then M < M N = G by maximality of M , so by
the Diamond Isomorphism Theorem, |G : M | = |N : M ∩ N |, which is a prime power
since N is elementary abelian (Exercise 31).

The previous two exercises show that the ‘top’ and ‘bottom’ of a finite solvable group
have a relatively simple structure.

Section 2

In the following exercises, for a group G, µ(G) denotes the minimal possible index of
a proper subgroup in G (cf. p. 203).

E XERCISE 1. Let G be a group, P ∈ Sylp (G), and suppose for all Q ∈ Sylp (G) with
P 6= Q, P ∩Q = 1. Then for all P 1 , P 2 ∈ Sylp (G) with P 1 6= P 2 , P 1 ∩P 2 = 1. Therefore the
number of nonidentity elements of p-power order in G is (|P | − 1)|G : NG (P )|.

Proof. Let P 1 , P 2 ∈ Sylp (G) with P 1 6= P 2 . By Sylow’s Theorem, write P 1 = g P g −1 and

P 2 = hPh −1 with g , h ∈ G and h −1 g 6∈ P . Now suppose towards a contradiction that
P 1 ∩ P 2 6= 1, so there exist x, y ∈ P with g xg −1 = h yh −1 6= 1. But then

1 6= (h −1 g )x(h −1 g )−1 = y ∈ P ∩ (h −1 g )P (h −1 g )−1

This contradicts that P has trivial intersection with distinct Sylow p-subgroups. Thus
P 1 ∩ P 2 = 1 as desired. The computational result follows trivially.

77
E XERCISE 4. There are no simple groups of order 80.

Proof. Suppose |G| = 80 = 24 · 5. By Sylow’s Theorem, n 5 ≡ 1(5) and n 5 |24 . If n 5 = 1,

G has a nontrivial proper normal Sylow 5-subgroup, so G is not simple. If n 5 6= 1,
then n 5 = 16, so the number of elements of order 5 is (5 − 1) · 16 = 64. This leaves
80 − 64 = 16 elements in G, hence we must have n 2 = 1, so again G is not simple.

E XERCISE 5. Let G be a solvable group with |G| = pm, p prime and (p, m) = 1. If
P ∈ Sylp (G) and NG (P ) = P , then G has a normal subgroup of order m.

Proof. Note n p = |G : NG (P )| = m, hence the number of elements of order p in G is

(p − 1) · m = pm − m. This leaves m remaining elements in G. Since G is solvable,
there exists M ≤ G with |M | = m (Hall). By Lagrange’s Theorem, M has no elements
of order p. Therefore M consists of precisely the remaining elements, and hence
M E G as desired.

E XERCISE 6. There are no simple groups of order 4125.

Proof. Suppose G is simple with |G| = 4125 = 3·53 ·11. Note µ(G) ≥ 15. Now n 5 ≡ 1(5),
n 5 |3 · 11, and n 5 > 1. But then n 5 = 11 < 15, a contradiction.

E XERCISE 8. There are no simple groups of order 792.

Proof. Suppose G is simple with |G| = 792 = 23 · 32 · 11. Then n 11 ≡ 1(11), n 11 |23 · 32 ,
and n 11 > 1, hence n 11 = 12. By simplicity then we may assume G ≤ S 12 . Now let
P ∈ Syl11 (G). Note (p. 204) |NS 12 (P )| = 11 · (11 − 1) = 2 · 5 · 11. But |NG (P )| = 2 · 3 · 11,
contradicting that NG (P ) ≤ NS 12 (P ) by Lagrange’s Theorem.

E XERCISE 10. There are no simple groups of order 4095.

Proof. Suppose G is simple with |G| = 4095 = 32 · 5 · 7 · 13. Then by computation,

n 13 = 3·5·7 and n 7 = 3·5. Let P ∈ Syl7 (G). Then |NG (P )| = 3·7·13. Note n 13 (NG (P )) = 1,
hence if Q ∈ Syl13 (NG (P )), then Q E NG (P ), so 3 · 7 · 13 divides |NG (Q)|. But also
Q ∈ Syl13 (G), hence |NG (Q)| = 3 · 13—a contradiction.

E XERCISE 14. There are no simple groups of order 144.

Proof. Suppose G is simple with |G| = 144 = 24 · 32 . Note µ(G) ≥ 6. By computation

then, we must have n 3 = 16. If Sylow 3-subgroups have pairwise trivial intersection,
then the number of nontrivial elements of 3-power order in G is (9 − 1) · 16, leaving
only 16 remaining elements in G and forcing n 2 = 1—a contradiction.
Therefore we may choose P,Q ∈ Syl3 (G) with |P ∩Q| = 3. Set R = P ∩ Q. Then
R E P and R E Q, hence P,Q ≤ NG (R). This implies that 32 divides |NG (R)| and
n 3 (NG (R)) > 1. Since n 3 (NG (R)) ≡ 1(3) and n 3 (NG (R)) divides 24 , n 3 (NG (R)) = 22
and hence 22 must also divide |NG (R)|. But then |G : NG (R)| divides 4, contradicting
the value of µ(G).

Section 3

78
E XERCISE 1. Let F (R) and F (S) be free groups. Then F (R) ∼
= F (S) iff |S| = |R|.

Proof. First suppose ϕ : S → R is a bijection. We may naturally regard this as an

injection ϕ : S → F (R), so by the universal property there exists Φ : F (S) → F (R) with
Φ|S = ϕ. Note that Φ is a bijection, so Φ witnesses as isomorphism F (S) ∼ = F (R).
Now suppose conversely Φ : F (S) ∼ = F (R). We merely sketch an idea: by methods
analogous to those in elementary vector space theory, we might be able to argue that
any minimal generating set of F (R) has the same cardinality as R, hence since S is
mapped bijectively to such a set under Φ, |S| = |R|.

E XERCISE 2. F (S) is abelian iff |S| = 1.

Proof. If |S| = 1, then F (S) is cyclic and hence abelian. If |S| > 1, choose a, b ∈ S with
a 6= b. Then by definition of products in F (S), [a, b] = a −1 b −1 ab 6= 1, hence a and b
do not commute, so F (S) is nonabelian.

E XERCISE 4. Every nonidentity element in F (S) has infinite order.

Proof. Again trivial by definition of products in F (S).

E XERCISE 5. We give a presentation for A 4 using two generators. Set σ = (123) and
τ = (124). We know from the lattice that A 4 = 〈σ, τ〉. Note σ3 = τ3 = (στ)2 = 1. The
latter relation implies that τσ = σ−1 τ−1 , so (as in the dihedral group) these relations
allow us to reduce any element of A 4 to a specific form σi τ j where 0 ≤ i , j < 3. Thus
any relations in A 4 may be derived from these three relations. It follows that

A 4 = 〈 σ, τ | σ3 = τ3 = (στ)2 = 1 〉

E XERCISE 11. Let S be a set and R = 〈 [s, t ] | s, t ∈ S 〉. Then A(S) = 〈S | R〉 is called the
free abelian group on S. If G is abelian and ϕ : S → G is any set map, then there exists
a unique homomorphism Φ : A(S) → G such that Φ|S = ϕ.
In particular, if |S| = n, then A(S) ∼
= n Z.
Q

Proof. Note we could simply define the homomorphism directly as in Theorem 17.
Alternately, note by the universal property of free groups that there exists a unique
Φ : F (S) → G with Φ|S = ϕ. Since G is abelian, R ≤ ker Φ, hence if N is the normal
closure of R, N ≤ ker Φ. Therefore we obtain a natural sequence

A(S) = F (S)/N −→ F (S)/ ker Φ −→ G

The composite homomorphism A(S) → G is ϕ on S, and is unique with this property.

If |S| = n, we can map the elements of S bijectively to the 1’s in the factors of n Z.
Q
Qn
The resulting homomorphism A(S) → Z is easily seen to be an isomorphism.

79