1

§1.4

Problem 8. If H and K are subgroups of finite index of a group G such that [G : H] and
[G : K] are relatively prime, then G = HK.

Proof. We have that [G : H ∩ K] = [G : H][H : H ∩ K] ≤ [G : H][G : K] with equality

iff [H : H ∩ K] = [G : K] iff G = HK by P4.8. Yet this is simple by our hypotheses; we
must have that [G : H], [G : K] | [G : H ∩ K] by T4.5, and since these indices are relatively
prime we have that [G : H ∩ K] ≥ [G : H][G : K]. This yields equality and completes the
proof.

Problem 11. Let G be a group of order 2n; then G contains an element of order 2. If n
is odd and G Abelian, there is only one element of order 2.

Proof. Let S consist of the unordered sets of elements in G such that the elements in each
pair are inverses of each other. We know that inverses are unique so that this collection S
is well-defined, that there is one singleton set - namely, {e} - and that the maximum size
of any set is two. Since |G| = 2n and every element has an inverse, it must be the case by
sheer counting that there is another singleton set in S - say {a}. Then a is its own inverse
and hence a has order 2. (This argument is due to Timbuc.)
If h1 , h2 are two distinct elements of G of order 2, then hh1 ihh2 i = {e, h1 , h2 , h1 h2 }, as
is clear after a routine check using the fact that G is Abelian. This is clearly a subgroup,
so |hh1 ihh2 i| = 4 | |G|, which contradicts the fact that n is odd. Hence the element of order
2 is unique in G.

Problem 12 will be taken for granted since it follows nearly directly from the proof of
P4.8.

Problem 13. If p > q are primes, a group of order pq has at most one subgroup of order
p.

Proof. If H, K are distinct subgroups of order p in a group G of order pq, then clearly
|H||K|
|H ∩K| = 1 by Lagrange, so that |H ∨K| ≤ |HK| = |H∩K| = p2 > pq = |G|, a contradiction
since H ∨ K < G.

§1.5

Problem 16. If f : G → H is a homomorphism, H is Abelian, and N is a subgroup of G

containing Ker f , then N / G.

Proof. We have that f¯ : G/ Ker f ∼ = Im f < H, so there exists an inverse isomorphism

f¯−1 . Note that f (N ) is normal in Im f since f (g)f (N )f (g −1 ) = f (g)f (N )f (g)−1 =
f (g)f (g)−1 f (N ) = f (N ) for g ∈ G (we used here that Im f < H is Abelian). Hence
gN g −1 = (g Ker f )N (g −1 Ker f ) = f¯−1 (f (g))N f¯−1 (f (g −1 )) = f¯−1 (f (gN g −1 )) = f¯−1 (N ) =
N Ker φ = N (using twice that N contains Ker f and that Ker f is a normal subgroup).

§1.8

Problem 4. Give an example to show that the weak direct product is not a coproduct in
the category of all groups.
2

Let us refresh on a few definitions:

Definition 1. A coproduct for {Ai }i∈I inside a category is an object of that category
together with morphism {ιi : Ai → S}i∈I such that the following property holds. For any
object B and family of morphisms {ψi : Ai → B}i∈I , there is a unique morphism ψ : S → B
such that ψ ◦ ιi = ψi for all i ∈ I.

This differs from a product in that a coproduct supplies “inclusion maps” into the object
of the coproduct, whereas the product supplies “projection maps” from the object of the
coproduct.

w
Q
Definition 2. The weak Q direct product of a family of groups of {Gi }i∈I , denoted { } i∈I Gi ,
is the set of all f ∈ i∈I Gi such that f (i) = ei for all but a finite number of i ∈ I. That
is, it is the set of all functions f : I → ∪i∈I Ai such that (i) f (i) ∈ Ai for all i, and (ii)
f (i) = ei ∈ Ai for all but finitely many i.

Proof. It suffices to show that the obvious inclusion maps of some subgroups H, K into the
weak direct product H × K - i.e., ιH (h) = (h, e) and ιK (k) = (e, k) are insufficient to satisfy
the definition. Using these two subgroups, we will also determine a group G with isomor-
phisms fH : H → G and fK : K → G for which there can be no homomorphism f : H ×K →
G such that f |H = fH and f |K = fK . Let G = H = K = S3 . Let fH , fG be the identity
maps. We will attempt to establish a homomorphism f : H × K → G that fulfills the re-
quirements above. Of course, we must mandate f ((h, e)) = fH (h) for h ∈ H and f ((e, k)) =
Definition of product
fK (k) for k ∈ K. Using these requirements, we find that f ((h, k)) =
Homomorphism
f ((h, e)(e, k)) = f ((h, e))f ((e, k)) = fH (h)fK (k). But we also have f ((h, k)) =
f ((e, k)(h, e)) = fK (k)fH (h), so that fH (h)fK (k) = fK (k)fH (h). Recalling that fH , fK are
identity maps, this gives us that hk = kh for any h ∈ H = S3 , k ∈ K = S3 . This implies
that S3 is Abelian, which is a contradiction.

Problem 8. Corollary 8.7 is false if one of the Ni is not normal.

Q
Proof. In general, the issue is that the product i∈I ai is no longer well-defined, since it
depends on the order of the ai . Consider G = D4 , with subgroups H and K where H is
the group of rotations and K is the group of flips. Clearly these subgroups generate G and
have trivial intersection, but G is no longer a direct product of these subgroups.

Problem 12. A normal subgroup H of a group G is said to a direct factor if there exists
a normal subgroup K of G such that G = H × K. (a) If H is a direct factor of K, and
K is a direct factor of G, then H is normal in G. (b) If H is a direct factor of G, then
every homomorphism H → G may be extended to an endomorphism G → G. However, a
monomorphism H → G need not be extendible to an automorphism G → G.

Proof. (a) We have that G = K × N2 = (H × N1 ) × N2 = H × N1 × N2 by the definition of

meaningful product. Hence if g = h0 n1 n2 ∈ G, we have gHg −1 = h0 n1 n2 Hn−1 −1 −1
2 n1 h0 =
−1
h0 Hh0 = H.
 3

(b) If φ : H → G = H × K is a homomorphism, let φ̄ : G → G be defined by

φ̄((h, k)) = φ(h). Then

φ̄((h1 , k1 )(h2 , k2 )) = φ̄((h1 h2 , k1 k2 ))

= φ(h1 h2 ) = φ(h1 )φ(h2 )
= φ̄((h1 , k2 ))φ̄((h2 , k2 )).

This proves the first part of this statement. Now note that the monomorphism f : Z2 →
S3 × Z2 given by 1 7→ (12) cannot be extended to an automorphism f¯ : S3 × Z2 → S3 × Z2 .
Such an automorphism would have to map ((12),0) to an element of order 2 such that
(12)f¯((12), 0) also has order 2, since (1, (12)) has order 2. The only options that satisfy the
first requirement are (13) and (23), and both generate 3-cycles, which do not satisfy the
second requirement.

§2.1
Problem 12. Let F be the free group on a set X and G the free group on a set Y . Let
F 0 < F be generated by {aba−1 b−1 | a, b ∈ F } and similarly for G0 . (a) F 0 / F , and F/F 0
is Abelian. (b) F/F 0 is a free Abelian group of rank |X|. (c) F ∼
= G iff |X| = |Y |.
Proof. (a) If g ∈ F , then gF 0 g = {gaba−1 b−1 } = {ga(g −1 g)b(g −1 g)a−1 (g −1 g)b−1 (g −1 g)b−1 g −1 } =
{(gag −1 )(gbg −1 )(gag −1 )−1 (gbg −1 )−1 } = F 0 , so F 0 is normal in F . Since (ab)(ba)−1 =
aba−1 b−1 ∈ F 0 , abF 0 = baF 0 , so F/F 0 is Abelian.
0 0 n
(b) We prove that {xF Pn | x ∈ X} is a basisPn for F/F . If0 y ∈PFn, then there0 exist {ai0}1 ⊂
n
Z, {xi }1 ⊂ X P such that i=1 ai xi = y, soP( i=1 ai xi ) + F = i=1 ai xi + F = y + F . Al-
n n
ternatively, if i=1 ai xi + F 0 = F P 0
, then i=1 ai xi ∈ F 0 . Hence there is an xyx−1 y −1 ∈ F 0
n
such that this sum is 0. Hence ( i=1 ai xi )xyx−1 y −1 = 0 ⇒ (ai xyx−1 y −1 ) = 0 for all i,
which implies that ai F 0 = F 0 for all i. This completes the proof.
(c) (⇐) is TI.7.8 by the definition of equivalence. (⇒) Let φ : F → G be an iso-
morphism and consider π : F → G/G0 , the projection map. Clearly π is surjective. If
f ∈ Ker π, then φ(f ) ∈ G0 ⇒ φ(f ) = g1 g2 g1−1 g2−1 for some g1 , g2 ∈ G. Since φ is surjective,
this means that φ(f ) = φ(f1 )φ(f2 )φ(f1−1 )φ(f2−1 ) = φ(f1 f2 f1−1 f2−1 ), and shice φ is injective,
f = f1 f2 f1−1 f2−1 , so f ∈ F 0 . The reverse implications clearly hold; if f1 f2 f1−1 f2−1 ∈ F , then
φ(f1 f2 f1−1 f2−1 ) = φ(f1 )φ(f2 )φ(f1 )−1 φ(f2 )−1 ∈ G0 = Ker π. So Ker π = F 0 , and the First
Isomorphism Theorem completes the proof.

§2.2
Problem 5. If G is a finitely generated Abelian group such that G/Gt has rank n and
H < G is such that H/Ht has rank m, then m ≥ n and (G/H)/(G/H)t has rank n − m.

Proof. It is worth proving that G ∼ = G/H ⊕ H if H is a normal subgroup of G. Let

G/H = {g1 H = eH, g2 H, . . . , gn H} where the sets are disjoint and {gi }n1 ⊂ G. Then every
element of g can be written uniquely as gi h for some i ∈ [n] and h ∈ H, and the obvious
map ensues.
We have G = G/Gt ⊕Gt . We have that G/Gt is finitely generated by T2.1 and Abelian,
and since it contains no torsion elements it is free. The same holds for H = H/Ht ⊕ Ht ; in
particular, H ∼= F , the maximal free group in H, so by identification H/Ht < H < G, so
H/Ht is a subgroup of the maximal free group of G, G/Gt . Hence by T1.6 m ≤ n.
4

(G/H)t is normal in G/H since G/H is Abelian. Assume for now that G/H is free
so that (G/H)t = {e}. We claim that if G is generated by X and H by Y , then since
G∼ = H ⊕G/H we must have that the rank of G/H is n−m. If G/H is not free, G/H/(G/H)t
is the maximal free subgroup in G/H, and this completes the proof.

Problem 7. Let G be P an Abelian torsion group. (a) G(p) is the unique maximum p-
subgroup of G. (b) G = G(p) where the sum is over all primes p such that G(p) 6= 0. (c)
If H is another Abelian torsion subgroup, then G ∼
= H iff G(p) ∼
= H(p) for all primes p.
Proof. (a) If H < G is a p-subgroup, then by definition it is contained in G(p). (b)
If g ∈ G is given and |g| = pn1 1 · · · · · pnmm , let mi = p|u|
ni . Then gcd({pi }) = 1, so by
i Pm
m
the extended Euclidean Algorithm there exist {ci }1 ⊂ Z such that i=1 ci mi = 1 ⇒
P m ni
i=1 ci mi u = u. Now |mi u| = pi , so |ci mi u| | |mi u| and ci mi u ∈ G(pi ). Therefore
G = G(2) + G(3) + · · · + G(p) + · · · , and since the G(p) are mutually disjoint and (trivially)
normal in G we get the claim.
(c) (⇒) Given φ : G → H, we can restrict φ to G(p) and land in H(p) since |φ(g)|
must divide the order of g for any group element. Similarly for the inverse function from
H(p) to G(p). If φp : G(p) → H(p) is given, then by (b) φ : G → H can be given by
(a1 , a2 , · · · ) 7→ (φ2 (a1 ), φ3 (a2 ), . . . ).

Problem 8. A finite abelian p-group G is generated by its elements of maximal order.

Proof. Since G is finite, it is finitely generated, and G ∼ = Zm1 ⊕ · · · ⊕ Zmn for m1 | · · · | mn .

Clearly each mn is a power of p since otherwise (0, . . . , 0, 1) does not have order a power
of p. Hence mi is a power of p for all i. We claim that X = {(z1 , . . . , zn ) | zn 6= 0}
generates the group. But this is simple; given arbitrary {zi }1n−1 , we have (z1 , . . . , zn−1 , 0) =
(z1 , . . . , zn−1 , 1) + (0, . . . , 0, mn − 1). Since X is exactly those elements of order mn , the
maximal order of the group, this completes the proof.

Problem 9. How many subgroup of order p2 does the Abelian group Zp3 ⊕ Zp2 have?

Proof. We start with the cyclic subgroups. If we choose the component of the generator in
Zp2 to have order p2 , then the subgroups generated by (kp, 1) for 0 ≤ k < p2 exhaust our
options. This gives p2 subgroups of order p2 . If we choose the component of the generator
in Zp2 to have order p, then (kp, p) for 0 ≤ k < p2 , gcd(k, p) = 1 exhaust our options. This
yields p − 1 subgroups; each one contains exactly p distinct elements of the form (kp, p)
with requirements of k given above. Finally we may choose 0 to be the component of the
generator in Zp2 , and (p, 0) gives another subgroup. This yields p2 +p cyclic subgroups. Any
non-cyclic subgroup of order p2 would have to have all elements of order 1 or p. There are 48
elements of order p in this group, and they together with the identity form one remaining
subgroup generated by (p, 0) and (0, p2 ). Hence the total number comes to p2 + p + 1
subgroups of order p2 .

Problem 15. Every finite subgroup of Q/Z is cyclic.

Proof. (Proof 1) Since S < Q/Z is finite, it has a smallest element s ∈ S (in the natural
order of Q). We claim that s generates the group. If there is an element t not in hsi,
then there exists a k such that ks < t < (k + 1)s (Archimedean Property). In particular,
0 < t − ks < s, so t − ks ∈ S and s is not the smallest element, a contradiction.
 5

Proof. (Proof 2, the way they wanted us to do it) P Let G be this group. Since G is finite,
it is a torsion group, and by Exercise 7(b) G = p prime G(p) where all but finitely many
of the summands are the identity. Note that G(p) consists of all the elements of G whose
reduced fractions have pn in the denominator; hence G(p) < Z(p∞ ). By Ex I.3.7(d), this
implies that G(p) is cyclic group of order a power of p. Hence G ∼ = Zpn1 1 ⊕ · · · ⊕ Zpnmm where
each prime is distinct. Since the gcd of any two of these indices is 1, we may by induction
show that this is congruent to Zpn1 1 ·····pnmm , which completes our claim.

§2.4

Problem 6. Let G be a group action on a set S containing at least two elements. Assume
that G is transitive. Prove (a) for x ∈ S, the orbit x̄ of x is S; (b) all the stabilizer Gx are
conjugate; (c) if G is faithful, N / G, and N < Gx for some x ∈ S, then N = hei; (d) for
x ∈ S, |S| = [G : Gx ]; hence |S| divides |G|.

Proof. (a) This follows from the definition of transitive. (b) Given x, y ∈ S, let g 0 ∈ G be
such that g 0 x = y. Then g 0 Gx g 0−1 y = g 0−1 Gx x = g 0−1 x = y. So g 0 Gx g 0−1 ⊂ Gy . Similarly,
g 0−1 Gy g 0 x = x, so g 0−1 Gy g 0 ⊂ Gx ⇒ Gy ⊂ gGx g −1 . (c) Since G is faithful, ∩x∈S Gx = {e},
so in particular if there is e 6= n ∈ N there is some y ∈ S such that n ∈ / Gy . Yet since Gx and
Gy are conjugate by (b), n cannot be in N . Hence N = {e}. (d) Let g1 Gx = eGx , . . . , gn Gx
be the distinct cosets of Gx in G. Then gi x 6= gj x for i 6= j (otherwise gi−1 g2 ∈ Gx ) Clearly
gi Gx x = gi x, and this exhausts all of the elements of G in the acting set. But G is transitive,
so {gi Gx x}n1 = S. Finally, note that |S| | [G : Gx ]|Gx | = |G|.

Problem 10. Show that the center of S4 is {e}; conclude that S4 is isomorphic to the group
of all inner automorphisms of S4 .

Proof. Any non-transitive non-trivial cycle is clearly not in C(S4 ) since it does not commute
with any cycle with elements both inside and outside the set of numbers that cycle oper-
ates on. Otherwise, (s1 s2 )(s3 s4 ) does not commute with (s2 s3 ), and (s1 s2 s3 s4 ) does not
commute with (s1 s2 ) since (s1 s2 )(s1 s2 s3 s4 )(s1 s2 )[s1 ] = s3 . The second statement follows
from the proof of C4.7(ii) since the homomorphism is given by conjugation and hence has
Inn(S4 ) as its image. The kernel is C(G) = {e}, so this is an isomorphism.

Problem 13. If a group G contains a proper subgroup of finite index, it contains a proper
normal subgroup of finite index.

Proof. Let [G : H] = n. Let G act on S, the set of cosets of H, by translation. This action
induces a homomorphism G → A(S), where |A(S)| = n! The kernel of this homomorphism
then has index at most n!.

Problem 14. If |G| = pn, p > n, p prime, and H is a subgroup of order p, then H / G.

Proof. Let G act on the set S of the n cosets of H by translation. This induces a ho-
momorphism φ : G → A(S) of order n! Note that p 6| n! since p > n is prime. Hence
| Ker φ|| Im φ| = |G| ⇒ | Ker φ|` = pn where ` | n! ⇒ ` = n ⇒ | Ker φ| = p. Since Ker φ < H
by P4.8, Ker φ = H and H is normal.

Problem 15. If a normal subgroup N of prime order p is contained in a group G of order

pn , then N is in the center of G.
6

Proof. Let G act on the elements of N by conjugation. This is a well-defined action since
N is normal in G. We claim this leads to a homomorphism φ from G to Aut(N ) where
g 7→ τg and τg (n) = gng −1 for n ∈ N . Clearly these are automorphisms on N since they are
injective homomorphisms (τg (n1 n2 ) = gn1 n2 g −1 = gn1 g −1 gn2 g −1 = τg (n1 )τg (n2 )). Now
any automorphism on N must send the identity to itself, and since N is cyclic there are p−1
possible automorphisms on N . But since | Ker φ|| Im φ| = pn and p is prime, | Im φ| = 1, so
every conjugation acts trivially on the elements of N and N < C(G).

§2.5

Problem 1. If N / G and N, G/N are both p-groups, then G is a p-group.

Proof. Let G → G/N be the canonical projection. Then any element of G maps to some
k
element of order a power of p - say g p - so this element is in the kernel of the projection -
k k ` k+`
i.e., g p ∈ N . Hence (g p )p = g p = e since N is a p-group, and since g ∈ G is arbitrary,
G is a p-group.

Problem 2. If G is a p-group, H / G, and H 6= {e}, then H ∩ C(G) 6= {e}.

Proof. First, if H is normal in G, then H = {ghg −1 h ∈ H} is the union of conjugacy classes

of G. Since each conjugacy class has order equal to the index of the centralizer of some
element in G, its order divides |G| = pn . Hence if |H| 6= 1, then p | |H|, and hence at least
p of the conjugacy classes in G making up H must have order 1 (since {e} is one of those
conjugacy classes) by the class equation. Yet any conjugacy class of order 1 has entries in
C(G), so H ∩C(G) 6= {e}. (Thanks to stackexchange user Arturo Magidin for the hint.)

Problem 3. Let |G| = pn . For each k, 0 ≤ k ≤ n, G has a normal subgroup of order pk .

Proof. This is clear for k = 0; we shall show an induction step and apply mathematical
induction. Let Hk be a normal subgroup of order pk in G. Then G/Hk is also a p-group
and hence has a non-trivial center, also a p-group. Let gk+1 Hk be an element of order p
in the center; then {gk+1 h : h ∈ Hk } =: Hk+1 is a subgroup of G of order pk+1 since H
commutes with gk+1 . Now for g ∈ G, g ∈ gHk , so it suffices to show that gHk is normal in
G/Hk . But this is clear since gk+1 Hk is in the center of G/Hk .

Problem 4. If G is an infinite p-group, then either G has a subgroup of order pn for each
n ≥ 1 or there exists m ∈ N ∪ {0} such that every finite subgroup of G has order ≤ pm .

Proof. This is equivalent to showing that every finite subgroup of order pn in G contains a
subgroup of order pn−1 . This is shown in the First Sylow Theorem.

Problem 9. If |G| = pn q, with p > q primes, then G contains a unique normal subgroup
of index q.

Proof. The existence of a (not necessarily normal) subgroup of index q is given by First
Sylow Theorem. Call this subgroup H. The number of Sylow p-subgroups must be (kp + 1)
for some k ∈ N ∪ {0} and must divide |G|. Since kp + 1 does not divide p for k > 0, it
would have to divide q; but kp + 1 > p > q, so this is impossible, k = 0, and by C5.8 H is
the unique Sylow p-subgroup.
 7

Problem 10. Every group of order 12, 28, 56, 200 must contain a normal Sylow subgroup
and hence is not simple.

Proof. The Third Sylow Theorem shows that the number of 7-Sylow subgroups or 5-Sylow
subgroups in any group of order 28 or 200 respectively must be unique; applying C5.8 yields
the statement. If there are 4 3-Sylow subgroups in a group of order 12, each have empty
intersection pairwise, so there are 8 elements of order 3 in this group, leaving three elements
which together with the identity must build the unique Sylow 2-subgroup. If there are 8
Sylow 7-subgroups in a group of order 56, each have empty intersection pairwise, so there
are 48 elements of order 7 in this group, leaving seven elements which together with the
identity must buidl the unique Sylow 2-subgroup. Third Sylow Theorem and C5.8 complete
the proof.

Problem 12. Show that every automorphism of S4 is an inner automorphism, and hence
S4 ∼
= Aut S4 .

Proof. Our finding that the center of S4 is the identity yielded that S4 = ∼ Inn S4 in Exercise
II.4.10, so the first part of the statement is all that is necessary to prove. Any automorphism
of S4 must permute the four Sylow 3-subgroups of S4 : h(123)i, h(124)i, h(134)i, h(234)i.
(There are no more by Third Sylow Theorem.) Note that any automorphism that fixes
h(123)i must fix 4 - similarly for the other three Sylow 3-subgroups of S4 . Hence if f, g are
automorphisms of S4 such that f (Pi ) = g(Pi ) for {Pi }41 the Sylow 3-subgroups of S4 , we
have g −1 f (Pi ) = Pi for all i ∈ [4], implying that g −1 f fixes each element of [4] and hence
that g −1 f is the identity ⇒ f = g. Therefore there are at most 24=4! automorphisms of S4
- those that permute the Sylow 3-subgroups - and since there are 24 inner automorphisms
of S4 and Inn S4 < Aut S4 we have Aut S4 = Inn S4 .

This leads to a problem interesting in its own right:

Problem. Any group of order 24 is not simple.

Proof. G has either 1 or 3 Sylow 2-subgroups. If it has 3, then conjugation sends these Sylow
2-subgroups to themselves, inducing an action of G on these subgroups Pi and further
inducing a homomorphism φ : G → S(Pi ) = S3 given by how G interacts with these
subgroups. The kernel of this homomorphism must have order ≥ 4 by order arguments.
But Ker φ / G.

The next few problems are from Artin’s Algebra, with solutions offered here as a way
to continue practicing group actions.

Problem. A nonabelian simple group cannot operate nontrivially on a set with fewer than
five elements.

Proof. Once it has been determined that the smallest nonabelian simple group is of order
60, the proof of this follows the same process as above. Here is another solution summarized
from a McGill’s professor’s website: it must be the case that the kernel of the homomorphism
G → S4 induced by this action on a set of four (or fewer) elements is trivial, so the
homomorphism is injective and we may consider G < S4 by inclusion. Now consider the
determinant map φ : G → {±}. If ker φ = G, then G is contained in A4 , and we have
8

already shown above that A4 is not simple. Orders 6 and 10 are products of two primes,
and orders 8 and 9 are powers of primes, exhausting all options for G; hence ker φ 6= G. If
ker φ = {e}, all nontrivial elements of G consist of odd permutations of order 2 and hence
is Abelian. This completes the proof.

Problem. If H is a proper subgroup of a finite group G, then G is not a union of conjugates

of H.

Proof. Let G act on NG (H) by conjugation. Now |NG (H)| ≥ |H|, so |G| = |OH ||GH | ⇒
|G| |G|
|O(H)| = |G H|
≤ |H| . Yet since every conjugate of H has at least a trivial intersection,
the union of the conjugates of H can have at most |O(H)|(|H| − 1) + 1 elements, which is
strictly less than |G| since H is proper.

Problem. Prove that no group of order 224 is simple.

Proof. 224 = 25 · 7. It is easily check that there are either 1 or 7 2-Sylow subgroups. If
there are seven, let φ : G → S7 be the induced homomorphism from the action of G on S7
by conjugation. Note that |S7 | = 5040 = 24 · 32 · 5 · 7; in particular, |G| 6| |S7 |, so φ is not
injective and Ker φ 6= {e}.

Problem. Let G be a group of order 30. (a) Prove that either the Sylow 5-subgroup K
of the Sylow 3-subgroup H is normal. (b) Prove that HK is a cyclic subgroup of G. (c)
Classify groups of order 30.

Proof. (a) Omitted; it is a Third Sylow Theorem argument. (b) Let H / G. Note that
(HK)5 = H(KHK)H(KHK)HK = H 5 K = K. Let φ : HK → K send g ∈ HK to g 5 .
Clear φ is not trivial - try letting h = e - so | Ker φ| = 5 and ∃hk ∈ HK such that h 6= e and
(hk)5 ∈ K \ {e}. We claim that this element has order 15. If |hk| = 3, then (hk)2 ∈ K \ {e}.
That is, hkhk ∈ K ⇒ hkh ∈ K. This shows that hkh = hh−1 k by the fact that H / G, so
hkh = k. Hence h is in the normalizer of K, and since h 6= e, |NG (K)| = 15 and K / G.
Hence H × K / G, and any element (h, k) where h, k 6= e is of order 15. (This is one possible
argument; can you find another?)
Now let K / G. Note (HK)3 = HK(HKH)K = HK 3 = H. Let φ : HK → H map
g 7→ g 3 . By similar reasoning to the above, ∃h ∈ H, k 6= e such that (hk)3 ∈ H \ {e}. If hk
has order 5, then (hk)6 = hk ∈ H \ {e}, so k = e, a contradiction. Hence |hk| = 15, and hk
generates G.
(c) By (b), we have that G = Z2 oφ Z15 where φ : Z2 → Aut(Z15 ) = Aut(Z3 ) ×
Aut(Z5 ) = Z2 × Z4 . Hence G is uniquely determined by where φ maps 1 ∈ Z2 . Since
φ(1) ∈ {(0, 0), (1, 0), (0, 2), (1, 2)}, these each classify the groups of order 30. Hence there
are at most four groups of order 30, and since Z2 ⊕ Z3 ⊕ Z5 , Z5 ⊕ S3 , Z3 ⊕ D5 , and D15 are
pairwise non-isomorphic, this completes the proof.

We return now to Hungerford. §2.6

Problem 1. Let G, H be groups, θ : H → Aut G a homomorphism. Let G oθ H be the set

G × H with the binary operation (g, h)(g 0 , h0 ) = (g[θ(h)(g 0 )], hh0 ). Show that G ×θ H is a
group.
 9

Proof. Closure is clear from definition. Associativity takes a bit of bookkeeping:

[(g, h)(g 0 , h0 )](g 00 , h00 ) = (gθ(h)(g 0 ), hh0 )(g 00 , h00 )

= (gθ(h)(g 0 )θ(hh0 )(g 00 ), hh0 h00 )
θ hom
= (gθ(h)(g 0 )θ(h)(θ(h0 )(g 00 )), hh0 h00 )
θ(h)∈Aut G
= (gθ(h)(g 0 θ(h0 )(g 00 )), hh0 h00 )
= (g, h)(g 0 θ(h0 )(g 00 ), h0 h00 )
= (g, h)[(g 0 , h0 )(g 00 , h00 )].

θ(h)∈Aut G
We note that (g, h)(e, e) = (gθ(h)(e), h) = (g, h) and (e, e)(g, h) = (θ(h)(g), h) =
θ(h−1 )∈Aut g
(g, h), yielding that (e, e) is the identity element of Goθ H. Finally, (θ(h−1 )(g −1 )θ(h−1 )(g), e) =
−1 −1 −1 −1 −1 θ hom −1 −1
(e, e) and (g, h)(θ(h )(g , h )) = (gθ(h)(θ(h (g )), e) = (e, e), yielding that (θ(h )(g ), h−1 )
and an inverse element of (g, h).

Problem 9. Classify up to isomorphism all group of order 18. Do the same for order 20.

Proof. We see that the Sylow 3-subgroup Sy3 must be normal in the group, so G = Sy3 oφ
Z2 . Since any group of order p2 for p prime is Abelian, Sy3 is one of two groups:

Case 1: Sy3 ∼ = Z9 . Then φ : Z2 → Aut(Z9 ) ∼ = (Z9 )× ∼

= Z2 × Z3 . Hence 1 7→ (0, 0) or
∼
1 7→ (1, 0). In the first, case G = Z2 × Z9 ; in the second, the order 2 elements conjugates
with the generator of Z9 by sending it to its inverse, and this group is easily seen to be D9 .

Case 2: Sy9 ∼ = Z3 × Z3 . First we note that Aut(Z3 × Z3 ) ∼ = GL2 (F3 ). Z3 × Z3 is generated

by (1, 0) and (0, 1). In any automorphism, (1, 0) may map to any of its eight non-identity
elements, and (0, 1) may map to anything outside of the cyclic group generated by that
element, yielding six choices for (0, 1). Hence there are 48 elements in Aut(Z3 × Z3 ). By
a similar line of reasoning, there are eight choices for the first column of M ∈ GL2 (F3 )
and six choices for the second column (columns must be linearly independent; that is, the
second column cannot be contained within the cyclic group generated by the first column),
so there are 48 elements in GL2 (F3 ). Finally, GL2 (F3 ) acts naturally on Z23 as a vector
space, inducing an isomorphism.
Any element of order 2 must have minimal polynomial (in the linear algebra sense) of
x2 − 1. Hence its eigenvalues can be either {1, −1} or {−1, −1}. (Thanks to stackexchange
user Derek Holt for this tip.) This leads naturally to the matrices which send generators
g, h of each copy of Z3 to the following: g 7→ g, h 7→ h (trivial), g 7→ g −1 , h 7→ h, or
g 7→ g −1 , h 7→ h−1 . The first yields G ∼ = Z2 × Z3 × Z3 , the second G ∼ = Z3 × S3 (note
the inverse flip under conjugation between the generators of Z2 and one copy of Z3 ), and a
group G ∼ = hx, y, z | |x| = 2, |y| = |z| = 3, xyx−1 = y −1 , xzx−1 zi.
This exhausts all cases, so these are the five groups of order 18.
Now for groups of order 20. The Sylow 5-subgroup is normal, so G ∼= Z5 oφ Sy2 . Again,
there are two cases for Sy2 :

Case 1: Sy2 ∼= Z4 . Then φ : Z4 → Aut(Z5 ) ∼

= Z4 can map 1 to the automorphism 1 7→ 1,
1 7→ 2, 1 7→ 3, and 1 7→ 4. We claim the middle two are isomorphic. This stems from the
10

fact that Z5 oφ17→2 Z4 ∼= hx, y | |x| = 4, |y| = 5, xyx−1 = y 2 i = hx−1 , y | |x−1 | = 4, |y| =
−1 3 ∼
5, x yx = y i = Z5 oφ17→3 Sy2 . The other two are neither isomorphic to each other - the
group generated from 1 7→ 1 is Abelian while 1 7→ 4 is not - nor to the group in question (x
in the group in question can only map to x and x−1 , y to y or y −1 , and none of these give
1 7→ 4), so this generates the groups Z5 × Z4 , G2 := hx, y | |x| = 4, |y| = 5, xyx−1 = y 2 i,
and G4 := hx, y | |x| = 4, |y| = 5, xyx−1 = y −1 i. It is easy to check that these are indeed
groups.
Case 2: Sy2 ∼ = Z2 × Z2 . Then φ : Z2 × Z2 → Aut(Z5 ) = ∼=∼ (Z5 )× ∼= Z4 can either map
trivially - in which case G ∼
= Z2 × Z2 × Z5 - or it can map one of the generators of Z2 × Z2
to the automorphism taking 1 7→ 4 in (Z5 )× and the other to the identity. This yields
Z5 × Z2 / G and the element of order 2 in the copy of Z2 that is not normal in G to act on
Z5 × Z2 by conjugation by sending elements to their inverse. This is D10 .
This exhausts all options, and hence the groups are Z4 × Z5 , Z2 × Z2 × Z5 , D10 , G2 ,
and G4 .
It is worth a note that there is one more map to consider in discussing possible maps
φ : Z2 × Z2 → (Z5 )× , which is the one that has both generators of Z2 × Z2 act on Z5 by
conjugation. We show that this case reduces to one of our other cases. Note that this would
generate the group hx, y, z | |x| = |y| = 2, |z| = 5, xzx−1 = z −1 = yzy −1 i. Since x = x−1 ,
this implies that z = xyz(xy)−1 and hence that z ∈ C(xy) ⇒ Z5 < N (hxyi) ⇒ hxyi / G
(since all of Z2 × Z2 is in the normalizer of this as well). Since |hxyi| = 2, we have that
Z5 × Z2 / G, so this reduces to the case of D10 .

Here are some interesting problems from previous A& M exams:

Problem 1. (J14.4, modified) Let R be a commutative ring. Let L, M, N, A be R-modules.
f g f∗
Suppose that 0 → L → M → N → 0 is a short exact sequence. Prove 0 → Hom(A, L) →
g∗ g∗ f∗
Hom(A, M ) → Hom(A, N ) and Hom(L, A) ← Hom(M, A) ← Hom(N, A) ← 0 are short
exact sequences where f ∗ , g ∗ are the contextually-derived induced homomorphisms of f, g,
respectively (for instance, if f : L → M , then f ∗ : Hom(A, L) → Hom(A, M ) is defined by
g 7→ f g and f ∗ : Hom(M, A) → Hom(L, A) is defined by g 7→ gf ).

Proof. We start with the first case. We want to show two things: f ∗ is injective, and
im f ∗ = ker g ∗ . Let f ∗ (φ) = f φ = f ψ = f ∗ (ψ). Then since f is injective, g = h. If
h ∈ im f ∗ , there exists φ ∈ Hom(A, L) such that f ∗ (φ) = f φ = h. Since im f = ker g, we
have gh = gf φ = 0, so h ∈ ker g ∗ and im f ∗ ⊂ ker g ∗ . If h ∈ ker g ∗ , then gh(x) = 0 for any
x. Hence h(x) is in the kernel of g, so for any element x ∈ M , there exists an element ` in
L such that f (`) = h(x) since im f = ker g. This element ` is unique since f is injective.
Define φ ∈ Hom(A, L) by g(x) = `. This is a homomorphism, and f g(x) = f (`) = h(x), so
h ∈ im f ∗ , and this completes the first part.
Now we consider the second case. We want to show two things: g ∗ is injective, and
im g = ker f ∗ . (This case is the dual of the first case, but there are interesting differences, so
∗

we include the proof.) If g ∗ (φ) = φg = ψg = g ∗ (ψ), then φ(x) = φ(g(yx )) for some yx ∈ M
since g is surjective, which in turn equals ψ(g(yx )) = ψ(x), so φ = ψ. If h ∈ im g ∗ , there
exists an element φ ∈ Hom(N, A) such that φg = h ⇒ hf = φgf = 0 since im g = ker f ,
so h ∈ ker f ∗ . If h ∈ ker f ∗ , then hf = 0, so 0 = h(ker f ) = h(im g). We can then
form h̄ : M/ ker g → A mapping m + ker f 7→ h(m), and we can also write the canonical
 11

isomorphism ḡ : M/ ker g → N ; this modding out by the kernel yields an isomorphism since
g is surjective. Hence for any n there is a unique element m+ker g such that f¯(m+ker g) = n.
Define ψ(n) = h(m). (Note that this is well-defined; this function is defined by composing
the inverse map of ḡ with f¯.) Then ψ is a homomorphism, and g ∗ (ψ) = ψg = h by definition
of ψ. So h ∈ im g ∗ and we are done.

Problem 2. Let R be an integral domain with field of fractions K, and let K̄ be an algebraic
closure of K. Fix α ∈ K̄. Suppose that M ⊂ K̄ is a finitely generated R-submodule such
that αM ⊂ M . Prove that there is a monic polynomial f ∈ R[x] such that f (α) = 0.

Proof. We have that det(A − Iα) = f (α) for some matrix A ∈ Mn×n (R) and f ∈ R[x]. We
will find a specific characteristic polynomial and then apply Cayley-Hamilton to complete
the proof. Let {miP }n1 be a set of generators for M . Then since αM ⊂ M , αmi ∈ M for all
n
i ∈ [n], so αmi = j=1 cij mj for any i ∈ [n]. Define A := [aij ]. Then A[mi ] = [αmi ] ⇒
(A − αI)[mi ] = 0. Hence det(A − αI) = f (α) = 0.

Problem 3. Let R be a commutative ring, and let M be a noetherian R-module. Set

I := {r ∈ R | ∀m ∈ M , rm = 0}. Prove that R/I is a noetherian ring.

Proof. Since the terms “noetherian ring” and “noetherian module” are not always defined
in some algebra courses, we define them quickly. A noetherian module is a module such
that every submodule is finitely generated iff it satisfies the ascending chain condition: if
A1 ⊂ A2 ⊂ · · · are modules, then there exists an n such that Am = An for all m ≥ n. A
noetherian ring is a ring such that every prime ideal is finitely generated iff it satisfies the
ascending chain condition (for ideals). Principal ideal domains (as rings and as modules
over themselves) are Noetherian.
Let {mi }n1 be a finite set of generators for M , and define a map R → M n by r 7→
→ Mn
(rm1 , . . . , rmn ). Then the kernel of this map is exactly the annihilator of M , so R/I ,−
n
as a submodule. Since M is finitely generated (naturally), R/I is finitely generated.

Problem 4. Let K be a field. Prove that the polynomial ring K[x] has infinitely many
maximal ideals.

Proof. This problem reduces to showing that K[x] has infinitely many irreducible elements,
even after multiplication by a unit. If K is infinite, (x + a) for a ∈ K is sufficient. If K
is finite, charK = p for p prime. If there are only finitely many irreducible polynomials
in K[x], then the algebraic closure of K is finite dimensional over K and hence is finite
of degree n. But then this algebraic closure contained the splitting field of the separable
n
polynomial xp − x over K, a contradiction. Since K[x] is a principal ideal domain, every
ideal of K[x] not contained in K contains one irreducible element, so there are infinitely
many maximal ideals in K[x]. (You can also use the theorem that for any n there is an
n
irreducible polynomial of degree n over a finite field - namely, xp − x.)

Problem 5. Let V be a finite dimensional vector space over C, and suppose we have C-
linear maps A1 , . . . , Ak : V → V such that for all i, j we have Ai ◦ Aj = Aj ◦ Ai . Show that
there exists a non-zero vector in V that is simultaneously an eigenvector for each of these
maps (with possible different eigenvalues).
12

Proof. We use induction on k. The base case is trivial. If there exists an element v which is
an eigenvector for Ai v for i ∈ [k − 1], note that Ai (Ak v) = Ak (Ai v) = Ak (λi v) = λi (Ak v),
so that multiplication by Ak preserves the joint eigenspace of {A1 , . . . , Ak−1 }. Now restrict
Ak to this joint eigenspace (well-defined by this preservation). Since C is algebraically
closed, we may consider this eigenspace as the set of linear combinations of some basis
of eigenvectors. This intersects non-trivially with the eigenspace of Ak restricted to this
joint eigenspace, and this subspace is as desired. Mathematical induction completes the
proof.

Problem 6. (Part (b) only) Let R be a commutative ring, and let P, F be left R-modules.
Assume P, F are finitely generated, that P is projective, and that F is free. Prove that
HomR (P, F ) is a projective R-module.

Proof. Since P is projective, F̃ ∼ = P ⊗ K for F̃ free and some R-module K and where F̃
can be chosen to be finitely generated (by the proof of this equivalence). We first claim
Hom(F̃ , F ) is free. If {fi }n1 forms a basis for F̃ , {gi }m
1 forms a basis for F , it is not hard to
show that {δij }n,m
i,j=1 , the Kronecker deltas, form a basis for this Hom group. Clearly each of
these exist in the Hom group. The rest follows since any element in Hom(F̃ , F ) is completely
and uniquely determined by how each fi maps to some fj for i ∈ [n] and j ∈ [m]. Then it
is easy to see that Hom(F̃ , F ) ∼ = Hom(P, F ) ⊕ Hom(K, F ) by the restriction maps.

Problem 7. Let R be an integral domain. Show that R is a field if and only if every
R-module is projective.

Proof. If R is a field, every R-module is (by definition) a vector space and is therefore free
and hence projective. If R is not a field, let I be a maximal ideal of R and consider R/I as
R/I
an R-module. We claim that following diagram cannot commute: h
id

R π R/I 0
for any h an R-module homomorphism. Since (0) 6= I, then for 0 6= a ∈ R, 0 6= i ∈ I,
π(a + i) = π(a). But then h must map a + I to both a + i and to a in R, contradiction.

Problem 8. Let k be a field, a ∈ k, and let p be a prime number. Prove that the polynomial
xp + a is either irreducible or has a root in k.

Proof. Let E be the splitting field of xp + a over k. Then xp + a = (x − z1 ) · · · · · (x − zp ) for

{z1 , . . . , zp } roots of the polynomial in E. If xp + a = f g forQf, g of degree ≥ 1, f factors
n
into (x − z1 ) · · · · · (x − zn ) for some n < p. We have z := 1 zi is equal to the additive
inverse of the constant term of f and hence is an element in k. Furthermore z p = (−a)n ,
since (z1 · · · · · zn )p = z1p · · · · · znp = (−a)n . Since n, p are relatively prime, there exist c, d such
that cn + dp = 1. Then (z c (−a)d )p = z cp (−a)dp = (−a)cn+dp = −a. This is an element in
k, so our polynomial has a root in k.
k k
Problem 9. Suppose z is any generator for the unit group of F4k . Prove that x2 +x+z 2 +z
has exactly 2k roots in F4k .
 13

Proof. First note that F4k has order a power of 2, so it is of characteristic 2. Hence
k k k
x2 + x + z 2 + z = (x + z)2 − (x + z) by Freshman’s Dream. This polynomial divides
2k 2k
(x+z)2 +(x+z); indeed, this polynomial is its square. Now the roots of (x+z)2 +(x+z)
2k
and x2 + x are clearly in bijective correspondence and are all contained in F4k , so this
k
former polynomial has 4k roots, so each factor of (x + z)2 + (x + z) must have 2k roots
exactly.

Problem 10. (b only) A group is said to be locally finite if every finitely generated subgroup
of the group is finite. Suppose that G is a group containing a normal subgroup K such that
K and G/K are both locally finite. Show that G is locally finite.

Proof. Map G → G/K in the canonical way. Any finitely generated subgroup H in G has
finitely many generators in its image through this map, so its index over K is finite. We use
the fact that every subgroup of finite index in a finitely generated subgroup is itself finitely
generated to get that, since [H : H ∩ K] < ∞ by this fact, H ∩ K is itself finitely generated
and is hence finite. Since |H| = [H : H ∩ K]|H ∩ K|, we are done.