Chapter 6

Soluble Groups
We have already met the concept of a composition series for a group. In this
chapter we shall consider groups whose composition factors are all abelian.
Think of this as the class of groups that can be built using only abelian
groups.

6.1

Definition of soluble groups

Definition 6.1 Let G be a group and x, y G. The commutator of x and y

is the element
[x, y] = x1 y 1 xy.
Note that the following equations hold immediately:
[x, y] = x1 xy
[x, y] = (y 1 )x y
xy = yx[x, y].

(6.1)

The latter tells us that the commutator essentially measures by how much
x and y fail to commute.
Lemma 6.2 Let G and H be groups, let : G H be a homomorphism
and let x, y, z G. Then
(i) [x, y]1 = [y, x];
(ii) [x, y] = [x, y];
(iii) [x, yz] = [x, z] [x, y]z ;
(iv) [xy, z] = [x, z]y [y, z].

62

Proof: (i) [x, y]1 = (x1 y 1 xy)1 = y 1 x1 yx = [y, x].

(ii) [x, y] = (x1 y 1 xy) = (x)1 (y)1 (x)(y) = [x, y].
(iii) For this and part (iv), we shall rely on Equation (6.1) and view it
as telling us how to exchange group elements at the expense of introducing
commutators. (This is known as collection.) So
xyz = yzx[x, yz]
but if we collect one term at a time we obtain
xyz = yx[x, y]z
= yxz[x, y]z
= yzx[x, z] [x, y]z .
Hence
yzx[x, yz] = yzx[x, z] [x, y]z ,
so
[x, yz] = [x, z] [x, y]z .
(iv)
xyz = zxy[xy, z]
and
xyz = xzy[y, z]
= zx[x, z]y[y, z]
= zxy[x, z]y [y, z].
Comparing we deduce
[xy, z] = [x, z]y [y, z].
!
Both parts (iii) and (iv) can be proved by a more simple-minded expansion of the terms on both sides, but more can be learnt and understood via
the collection process.
Definition 6.3 Let G be a group. The derived subgroup (or commutator
subgroup) G" of G is the subgroup generated by all commutators of elements
from G:
G" = # [x, y] | x, y G $.
Part (i) of Lemma 6.2 tells us that the inverse of a commutator is again
a commutator, but we have no information about products of commutators.
Consequently, a typical element of G" has the form
[x1 , y1 ] [x2 , y2 ] . . . [xn , yn ]
where xi , yi G for each i.
63

Example 6.4

(i) In an abelian group G

[x, y] = x1 y 1 xy = 1

for all x and y,

so G" = 1. (The condition G" = 1 is equivalent to G being abelian.)

(ii) Lets calculate the derived group of Q8 . Note that 1 is central in Q8 ,
so we only need to calculate the commutators of i, j and k.
[i, j] = (i)(j)ij = ij(ij) = (ij)k = k2 = 1
[i, k] = (i)(k)ik = i(ki)k = i(jk) = i2 = 1
[j, k] = (j)(k)jk = jk(jk) = (jk)i = i2 = 1
So Q"8 = #1$.
Definition 6.5 The derived series (G(i) ) (for i " 0) is the chain of subgroups of the group G defined by
G(0) = G
and
G(i+1) = (G(i) )"

for i " 0.

So G(1) = G" , G(2) = (G" )" = G"" , etc. We then have a chain of subgroups
G = G(0) " G(1) " G(2) " .
We shall see later that this is indeed a series in the sense of Definition 4.1
(in that each term is normal in the previous).
Definition 6.6 A group G is soluble (solvable in the U.S.) if G(d) = 1 for
some d. The least such d is the derived length of G.
Since when forming the derived series, we take the derived subgroup of
the previous term at each stage, once we have a repetition then the series
becomes constant. Thus if G is a soluble group of derived length d, its
derived series has the form
G = G(0) > G(1) > G(2) > > G(d) = 1.
Example 6.7 Looking back at Example 6.4 we see that:
(i) Any abelian group is soluble with derived length 1.
(ii) The group Q8 is soluble with derived length 2.
We seek to understand the properties of soluble groups, and to produce
equivalent formulations so that examples can be more easily described. Accordingly, we begin by establishing basic properties of the derived subgroup
and the derived series.
64

Lemma 6.8

(i) If H is a subgroup of G, then H " # G" .

(ii) If : G K is a homomorphism, then G" # K " .

(iii) If : G K is a surjective homomorphism, then G" = K " .
Proof: (i) If x, y H, then [x, y] is a commutator of elements of G so
belongs to the derived subgroup of G:
[x, y] G"

for all x, y H.

Therefore
# [x, y] | x, y H $ # G" ,
so H " # G" .
(ii) If x, y G, then [x, y] = [x, y] K " . Since K " is closed under
products, it follows that any product of commutators in G is mapped into K "
by . Thus G" # K " .
(iii) Let a, b K. Since is surjective, there exists x, y G such that
a = x and b = y. Thus
[a, b] = [x, y] = [x, y] G" .
Thus
[a, b] G"

for all a, b K.

This forces K " # G" . Using (ii) gives K " = G" , as required.

Lemma 6.9 Subgroups of soluble groups are soluble.

Proof: Let G be a soluble group and H be a subgroup of G. We start by
claiming that H (i) # G(i) for all i. We will prove the claim by induction
on i. The case i = 0 is the inclusion H # G which holds by assumption.
Now suppose H (i) # G(i) . Apply Lemma 6.8(i) to give
(H (i) )" # (G(i) )" ;
that is,
H (i+1) # G(i+1) .
This completes the induction.
Now since G is soluble, G(d) = 1 for some d. Therefore, as H (d) # G(d) ,
we have H (d) = 1 and so we deduce that H is soluble.
!
Lemma 6.10 Homomorphic images of soluble groups are soluble.

65

Proof: Let G be a soluble group, and let K be a homomorphic image of G.

Thus there exists a surjective homomorphism : G K. We claim that
K (i) = G(i) for all i. We will prove the claim by induction on i. The case
i = 0 is the equation K = G which holds by assumption.
Now suppose K (i) = G(i) . Thus induces a surjective homomorphism
(i)
G K (i) and Lemma 6.8(iii) gives
(K (i) )" = (G(i) )" ;
that is,
K (i+1) = G(i+1) .
This completes the induction.
Now as G is soluble, G(d) = 1 and thus
K (d) = G(d) = 1 = 1.
Hence K is soluble.

It follows that quotient groups (which are the same as homomorphic

images) of soluble groups are themselves soluble. There is a rather strong
converse to the above lemma:
Proposition 6.11 Let G be a group and N be a normal subgroup of G
such that both G/N and N are soluble. Then G is soluble.
Proof: Let : G G/N be the natural map. By assumption (G/N )(d) =
1 and N (e) = 1 for some d and e. Now, by Lemma 6.10,
G(d) = (G/N )(d) = 1.
Hence
G(d) # ker = N.
Therefore, by Lemma 6.9,
(G(d) )(e) # N (e) = 1;
that is,
G(d+e) = 1.
Thus G is soluble.

66

Note It is not necessarily the case that G has derived length d + e: this
is just an upper bound. For example, if G is the direct product of N and
G/N then one can show that the derived length of G is the maximum of d
and e.
We have observed that if : G K is a surjective homomorphism then
G" = K " . In particular, if is an automorphism of G (that is, an isomorphism G G), then G" = G" . We give the following special name to
subgroups satisfying this property.
Definition 6.12 A subgroup H of a group G is said to be a characteristic
subgroup of G if x H for all x H and all automorphisms of G. We
write
H char G
to indicate that H is a characteristic subgroup of G.
The definition requires that H # H for all automorphisms of G. But
then H1 # H, and applying yields H # H. Thus H is a characteristic
subgroup if and only if H = H for all automorphisms of G.
Example 6.13 (i) The trivial subgroup 1 is fixed by all automorphisms
of G, and hence is characteristic in G.
(ii) The group G is a characteristic subgroup of itself, since all automorphisms are bijections.
(iii) Consider the group S5 . The only normal subgroups of S5 are 1, A5 ,
and S5 . Since A5 is the only normal subgroup of S5 of order 60, it
must be fixed by all automorphisms of S5 , so A5 char S5 .
(iv) Let G = C15 = C3 C5 . Then the subgroup of order 3 in G contains
all elements of order 3 in G, and hence must be mapped to itself by
all automorphisms of G. The same applies to the subgroup of order 5.
Thus both C3 and C5 are characteristic subgroups of G.
(v) Consider the group V4 = {1, a, b, c}. This has three proper nontrivial
normal subgroups, all of order 2, generated by a, b and c respectively.
The permutation (a b c) is an automorphism of V4 . Hence, no proper
nontrivial normal subgroup of V4 is fixed by all automorphisms of V4 ,
and hence the only characteristic subgroups of V4 are 1 and V4 .
Our observation above then is that
G" char G
for all groups G and we shall soon see that all terms in the derived series
are also characteristic.
67

Lemma 6.14 Let G be a group.

(i) If H char G, then H ! G.
(ii) If K char H and H char G, then K char G.
(iii) If K char H and H ! G, then K ! G.
Thus there is considerable difference between characteristic subgroups
and normal subgroups. For example, note that in general
K ! H ! G does not imply K ! G.
If : G K is a homomorphism and H char G, then it does not follow necessarily that H char G. (Consequently the Correspondence
Theorem does not work well with characteristic subgroups.)
If H # L # G and H char G, then it does not necessarily follow that
H char L.
Proof of Lemma 6.14: (i) If x G, then x : g ' gx is an automorphism
of G. Hence if H char G, then
H x = Hx = H

for all x G,

so H ! G.
(ii) Let be an automorphism of G. Then H = H (as H char G). Hence
the restriction |H of to H is an automorphism of H and we deduce
x K

for all x K

(since this is the effect that the restriction |H has when applied to elements
of K). Thus K char G.
(iii) Let x G. Then H x = H (as H ! G) and therefore x : g ' gx (for
g H) is a bijective homomorphism H H; that is, x is an automorphism
of H. Since K char H, we deduce that K x = Kx = K. Thus K ! G.
!
We have seen that G" char G. Recall the definition of the derived series:
G(0) = G,

G(i+1) = (G(i) )"

for i " 0.

Therefore
G(i) char G(i1) char G(i2) char char G(1) char G(0) = G.
Applying Lemma 6.14(ii) we see that each G(i) is a characteristic subgroup
(and hence a normal subgroup) of G for each i.

68

Proposition 6.15 The derived series

G = G(0) " G(1) " G(2) "
is a chain of subgroups, each of which is a characteristic subgroup of G, and
hence each of which is a normal subgroup of G.
!
In particular, if G is a soluble group of derived length d then
G = G(0) > G(1) > > G(d) = 1
and this is a normal series (each term is normal in G). In particular, we can
consider the factors
G(0) /G(1) , G(1) /G(2) , . . . , G(d1) /G(d) ;
i.e., the quotient groups G(i) /(G(i) )" for i = 0, 1, . . . , d 1.
We now deduce some information about these factors.
Lemma 6.16 Let G be a group and N be a normal subgroup of G. Then
G/N is abelian if and only if G" # N .
In particular, G/G" is an abelian group and it is the largest quotient
group of G which is abelian. We often call G/G" the abelianisation of G.
Proof: Suppose G/N is abelian. Then
Nx Ny = Ny Nx

for all x, y G,

so
N [x, y] = (N x)1 (N y)1 (N x)(N y) = N 1

for all x, y G.

Thus [x, y] N for all x, y G and we obtain G" # N .

Conversely if G" # N , then [x, y] N for all x, y G and reversing the
above steps shows that G/N is abelian.
!
In particular, the factors occurring in the derived series are all abelian.
So if G is a soluble group, then its derived series is a normal series with all
factors abelian. The following result strengthens this.
Theorem 6.17 Let G be a group. The following conditions are equivalent:
(i) G is soluble;
(ii) G has a chain of subgroups
G = G0 " G1 " G2 " " Gn = 1
such that Gi is a normal subgroup of G and Gi1 /Gi is abelian for
i = 1, 2, . . . , n;
69

(iii) G has a chain of subgroups

G = G0 " G1 " G2 " " Gn = 1
such that Gi is a normal subgroup of Gi1 and Gi1 /Gi is abelian for
i = 1, 2, . . . , n.
Condition (ii) says that G has a normal series with abelian factors, while
(iii) says that G has a subnormal series with abelian factors.
Proof: (i) (ii): The derived series is such a chain of subgroups.
(ii) (iii): Immediate: If Gi ! G for Gi # Gi1 # G, then Gi ! Gi1 .
(iii) (i): Suppose
G = G0 " G1 " G2 " . . . " Gn = 1
is a series where Gi1 /Gi is abelian for all i.
Claim: G(i) # Gi for all i.
We prove the claim by induction on i. Since G(0) = G = G0 , the claim
holds for i = 0.
Suppose G(i) # Gi . Now Gi+1 ! Gi and Gi /Gi+1 is abelian. Hence
(Gi )" # Gi+1 by Lemma 6.16. Further, by Lemma 6.8(i), (G(i) )" # (Gi )" .
Hence
G(i+1) = (G(i) )" # (Gi )" # Gi+1 .
Hence by induction G(n) # Gn = 1, so G(n) = 1 and G is soluble.

We now have a characterisation that a group is soluble if and only if it

has a series with abelian factors. We shall obtain a further such equivalence
by linking solubility to composition series.
We saw in Example 6.7 that all abelian groups are soluble. In particular,
the infinite cyclic group is soluble, though we know (Example 4.4) that this
group does not have a composition series. Accordingly we cannot hope for
composition series to give us complete information about soluble groups. It
turns out that as long as we avoid the infinite soluble groups, composition
series do tell us whether or not our group is soluble.
Theorem 6.18 Let G be a group. The following conditions are equivalent:
(i) G is a finite soluble group;
(ii) G has a composition series with all composition factors cyclic of prime
order.
Recall that the abelian simple groups are precisely the cyclic groups of
(various) prime orders. Thus part (ii) describes the groups with abelian
composition factors.
70

Proof: (ii) (i): Let

G = G0 > G1 > G2 > > Gn = 1
be a composition series for G and suppose that all of the factors are cyclic.
Then Gi ! Gi1 and Gi1 /Gi is abelian for each i. Thus this is a chain of
subgroups as in Theorem 6.17(iii) and therefore G is soluble. Further
|G| = |G0 /G1 | |G1 /G2 | . . . |Gn1 /Gn |,
a product of finitely many primes, so G is finite.
(i) (ii): Let G be a finite soluble group. Then by Theorem 6.17,
G possesses a chain of subgroups
G = G0 > G1 > G2 > > Gn = 1

(6.2)

such that Gi ! Gi1 and Gi1 /Gi is abelian for all i. Note that G can only
have at most finitely many such series. Thus we may assume that (6.2) is
the longest series for G with abelian factors. Such a series must then be
a composition series: for if some Gi1 /Gi is not simple, then there exists
N ! Gi1 with Gi < N < Gi1 . We then obtain a series
G = G0 > > Gi1 > N > Gi > > Gn = 1
which is longer than (6.2) and the new factors occurring here satisfy
N/Gi # Gi1 /Gi

and

Gi1 /Gi
Gi1 /N
=
N/Gi

(by the Third Isomorphism Theorem). Since Gi1 /Gi is abelian, we see
that N/Gi and Gi1 /N are abelian. This contradicts the assumption that
(6.2) is the longest series with abelian factors.
We now deduce that (6.2) is indeed a composition series and hence the
composition factors of G are abelian. Since the only abelian simple groups
are cyclic of prime order, we deduce that all the composition factors of G
are cyclic of prime order (for various primes).
!
Example 6.19 In Example 4.3 we saw that
S4 > A4 > V4 > #(1 2)(3 4)$ > 1
is a composition series for S4 and the composition factors are C2 , C3 , C2
and C2 . Hence S4 is soluble by Theorem 6.18.
Example 6.20 The dihedral group D2n contains an element of order n,
so #$ has index 2, and so is normal. Thus
D2n > #$ > 1
is a series for D2n with both factors cyclic. Hence D2n is soluble by Theorem 6.17.
71

Example 6.21 If n 5, then An is non-abelian simple group, so is insoluble

by Theorem 6.18. As every subgroup of a soluble group is soluble, it follows
that Sn is insoluble for n 5.
Careful analysis of the examples in Section 3 shows that the groups we
considered that were not simple in 3.73.10 are also soluble groups.

6.2

Finite soluble groups

For the rest of this chapter we shall work only with finite groups. Our goal
is to prove Halls Theorem on finite soluble groups.

6.2.1

Minimal normal subgroups

In this subsection we will work with finite groups without assuming that
they are also soluble.
Definition 6.22 Let G be a finite group. A minimal normal subgroup of G
is a non-trivial normal subgroup of G which properly contains no nontrivial
normal subgroups of G.
Thus M is a minimal normal subgroup of G if
(i) 1 < M ! G;
(ii) if 1 # N # M and N ! G, then either N = 1 or N = M .
Note that, apart from the trivial group, all finite groups have minimal
normal subgroups. To see this, we start with the group G itself. If this isnt
a minimal normal subgroup, then there is a nontrivial subgroup below it
which is normal. If this isnt minimal, then there is a nontrivial subgroup
below it which is normal in G. Repeating this process must eventually stop
(since G is finite) and yield a minimal normal subgroup.
Example 6.23 (i) If n 5, then An is simple. Thus there are no normal
subgroups of Sn contained in An , and hence An is a minimal normal
subgroup of Sn .
(ii) Let G = A5 A6 . Then the normal subgroups of G are 1, A5 , A6 and
G. Therefore the minimal normal subgroups of G are A5 and A6 .
Definition 6.24 A non-trivial group G is called characteristically simple if
the only characteristic subgroups it has are 1 and G.
(Recall, from Definition 6.12, that a characteristic subgroup of G is a
subgroup which is closed under applying all automorphisms of G.)

72

Example 6.25 (i) If G is a simple group, then 1 and G are the only
normal subgroups of G, so G is characteristically simple.
(ii) Recall Example 6.13(v). We proved that V4 has no characteristic subgroups, so V4 is characteristially simple.
(iii) If G is nonabelian and soluble, then G" char G and 1 < G" < G, so G
is not characteristically simple.
The following lemma is easy to prove, but often extremely useful.
Lemma 6.26 Let G be a group, and let M be a minimal normal subgroup
of G. Then M is characteristically simple.
Proof: Let K be a characteristic subgroup of M . Then
K char M ! G,
so K ! G by Lemma 6.14(iii). Thus minimality of M forces K = 1 or
K = M . Hence M is indeed characteristically simple.
!
Theorem 6.27 A characteristically simple finite group is a direct product
of isomorphic simple groups.
Proof: Let G be a finite group which is characteristically simple. Let S be
a minimal normal subgroup of G. (So S += 1. It is possible that S = G.)
Consider the following set
D = { N ! G | N = S1 S2 Sk where each Si is a
minimal normal subgroup of G isomorphic to S }.
(Recall what we mean by the direct product here: it is an internal direct
product, so we need Si S1 . . . Si1 Si+1 . . . Sk = 1 for each i, as well as
N = S1 S2 . . . Sk . We already assume Si ! G, so the requirement Si ! N
comes for free.)
Note that S D, so D certainly contains non-trivial members. Choose
N D of largest possible order. We will prove that N = G.
If N += G then, as G is characteristically simple, N is not characteristic
in G. Hence there exists an automorphism of G such that
N ! N.
Let N = S1 S2 Sk , then there exists i such that
Si ! N.

73

Now is an automorphism of G, so Si is a minimal normal subgroup of G.

Now N Si ! G and N Si is properly contained in Si (as Si ! N ).
Therefore, by minimality, N Si = 1. It follows that
N Si = N Si = S1 S2 Sk Si
and
N Si ! G.
This shows that N Si D. This contradicts N being a maximal member
of D.
Therefore
G = N = S1 S2 Sk ,
where each Si is a minimal normal subgroup of G isomorphic to our original
minimal normal subgroup S.
It remains to check that S is simple. If J ! S1 , then
J ! S1 S2 Sk = G.
Therefore, as S1 is a minimal normal subgroup of G, either J = 1 or J = S1 .
Hence S1 (and accordingly S) is simple.
Thus the result follows.
!
The following key theorem now follows easily:
Theorem 6.28 A minimal normal subgroup of a finite group G is a direct
product of isomorphic simple groups.
Proof: By Lemma 6.26, a minimal normal subgroup N of G is characteristically simple. By Theorem 6.27, the group N is therefore a direct product
of isomorphic simple groups.
!
In a finite soluble group, a minimal normal subgroup is therefore a direct
product of cyclic groups of order p (for some prime p).
Definition 6.29 Suppose that p is a prime number. An elementary abelian
p-group G is an abelian group such that
xp = 1

for all x G.

Recall that a finite abelian group is a direct product of cyclic groups. It

follows that a finite group is an elementary abelian p-group if and only if
G
= Cp Cp Cp
"#
$
!
d times

for some d.
Putting together Theorem 6.18 and Theorem 6.28 gives:
74

Theorem 6.30 A minimal normal subgroup of a finite soluble group is an

elementary abelian p-group for some prime number p.
!
Example 6.31 (i) The group S4 has minimal normal subgroup V4 : we
have already noted that V4 is characteristically simple (Example 6.25).
It is clear that V4 is an elementary abelian 2-group.
(ii) Let G = Q8 . Then all proper normal subgroups of G contain 1, so
#1$ is the unique minimal normal subgroup of G. Since #1$
= C2 ,
it is clear that it is an elementary abelian 2-group.

6.2.2

Hall subgroups

Definition 6.32 Let be a set of prime numbers and let G be a finite

group. A Hall -subgroup of G is a subgroup H of G such that |H| is a
product involving only the primes in and |G : H| is a product involving
only primes not in . A subgroup of G is called a -subgroup if its order is
a product involving only the primes in .
If p is a prime number, then a Hall {p}-subgroup is precisely the same
thing as a Sylow p-subgroup.
Example 6.33 Consider the alternating group A5 of degree 5. Here
|A5 | = 60 = 22 3 5.
So a Hall {2, 3}-subgroup of A5 has order 12. We already know of a subgroup
with this order: thus, A4 is a Hall {2, 3}-subgroup of A5 .
A Hall {2, 5}-subgroup of A5 would have order 20 and index 3, while
a Hall {3, 5}-subgroup of A5 would have order 15 and index 4. If H were
one of these, then we could let A5 act on the cosets of H and obtain a
homomorphism : A5 Sr (where r = 3 or 4). Here ker += 1 and ker +=
A5 (as ker # H), which would contradict the fact that A5 is simple.
Hence A5 does not have any Hall {2, 5}-subgroups or Hall {3, 5}-subgroups.
So in insoluble groups, some Hall -subgroups might exist, while others
might not (in fact, it is a theorem that some definitely do not!). This is in
stark contrast to soluble groups:
Theorem 6.34 (P. Hall, 1928) Let G be a finite soluble group and let
be a set of prime numbers. Then
(i) G has a Hall -subgroup;
(ii) any two Hall -subgroups of G are conjugate;
(iii) any -subgroup of G is contained in a Hall -subgroup.
75

There is a clear analogy between this and Sylows Theorem (3.4).

Hall subgroups and this theorem are named after Philip Hall (1904
1982), a British mathematician who did groundbreaking research into the
theory of finite and infinite groups in the early and mid-parts of the twentieth
century.
A number of tools are needed in the course of this theorem. The one
remaining fact that has not already been established is the following result.
Lemma 6.35 (Frattini Argument) Let G be a finite group, N be a normal subgroup of G and P be a Sylow p-subgroup of N . Then
G = NG (P ) N.
The name of the lemma suggests (correctly) that it is the method of
proof that is actually most important here. The idea can be adapted to
many situations and turns out to be very useful.
Proof: Let x G. Since N ! G,
P x # N x = N,
so P x is a Sylow p-subgroup of N . Sylows Theorem then tells us that
P x and P are conjugate in N :
Px = Pn

for some n N .

Therefore
P xn

= P,

so y = xn1 NG (P ). Hence x = yn NG (P )N . The reverse inclusion is

obvious, so
G = NG (P )N.
!
Proof of Theorem 6.34: Our strategy is to prove part (i) and deduce
part (iii) by showing that a -subgroup is contained in a conjugate of the
Hall -subgroup found already. We will then deduce part (ii) at the end.
We will prove the following by induction on the order of G:
G has a Hall -subgroup H;
if L is a -subgroup of G then L is contained in some conjugate of H.
Both are trivial if |G| = 1. Assume then that |G| > 1 and that these
statements hold for soluble groups of order smaller than G. Write |G| = mn
where m is a product involving primes in and n is a product involving
primes not in . (A Hall -subgroup of G is then a subgroup of order m.)
We can assume that m > 1 since otherwise the statements are trivially true.
Let M be a minimal normal subgroup of G. By Theorem 6.30, M is
elementary abelian. We consider two cases according to the prime dividing
the order of M .
76

Case 1:
Then

M is an elementary abelian p-group where p . Write |M | = p .

|G/M | = mn/p = m1 n,

where m = m1 p . By induction, the above statements hold for G/M . The

Correspondence Theorem tells us that a Hall -subgroup of G/M has the
form H/M where H is a subgroup of G containing M . Then
|H/M | = m1
so
|H| = m1 |M | = m1 p = m.
Hence H is a Hall -subgroup of G.
Now let L be any -subgroup of G. The image LM/M (
= L/(L M ))
of L in the quotient group is a -subgroup of G/M . Hence, by induction,
some conjugate of H/M contains LM/M , say
LM/M # (H/M )M x = H x /M
where x G. Thus
L # LM # H x .
This completes Case 1.
Case 2: No minimal normal subgroup of G is an elementary abelian
p-group with p . In particular, our minimal normal subgroup M of G
satisfies |M | = q where q
/ . Then
|G/M | = mn/q = mn1
where n = n1 q . We now further subdivide according to n1 .
Subcase 2A: n1 += 1.
By induction, G/M has a Hall -subgroup, which has the form K/M
where K is a subgroup of G containing M and
|K/M | = m.
Then
|K| = m|M | = mq = mn/n1 < mn.
Now, |K| < |G| and hence by induction K possesses a Hall -subgroup. Let
H be a Hall -subgroup of K. Then |H| = m, so H is also a Hall -subgroup
of G.
Now let L be a -subgroup of G. The image LM/M of L in the quotient
group is a -subgroup of G/M . Hence, by induction, LM/M is contained
in some conjugate of K/M ; say
LM/M # (K/M )M x = K x /M
77

where x G. Hence L # LM # K x , so Lx # K. Then Lx is a

1
-subgroup of K and since |K| < |G|, we deduce by induction that Lx #
H y for some y K. Hence
L # H yx
and we have completed Subcase 2A.
Subcase 2B: n1 = 1, so |G| = mq .
Note also that the general assumption of Case 2 still applies: G has no
minimal normal subgroup which is an elementary abelian p-group for p .
Now |G/M | = m > 1. Let N/M be a minimal normal subgroup of G/M .
Then N/M is an elementary abelian p-group for some p (since m is a
product involving only primes in ), say |N/M | = p . Then N ! G and
|N | = p q .
Let P be a Sylow p-subgroup of N . Let us now apply the Frattini Argument
(Lemma 6.35):
G = NG (P )N.
But N = P M , so
G = NG (P )P M = NG (P )M
(as P # NG (P )).
Now consider J = NG (P ) M . We will prove that J = 1. Since M is
abelian, J ! M . Also since M ! G, J = NG (P ) M ! NG (P ). Hence
J ! NG (P )M = G.
But M is a minimal normal subgroup of G, so J = 1 or J = M . If J =
NG (P ) M = M , then M # NG (P ), so G = NG (P ). Hence P is a normal
p-subgroup of G and some subgroup of P is a minimal normal subgroup of G
and this is then an elementary abelian p-group with p . This is contrary
to the general assumption made for Case 2.
Thus J = 1, so NG (P ) M = 1. Now
mq = |G| = |NG (P )M | = |NG (P )| |M |,
so |NG (P )| = m. Hence H = NG (P ) is our Hall -subgroup. (We have now
completed part (i) of the theorem!)
Now consider some -subgroup L of G. We have shown already that
G = HM , so
LM = LM G
= LM HM
= (LM H)M
78

!
!"
"!HM
!
LM !
!"
"
"
!
"!
"!H
!
!"
!
!
"!
M!
! LM H
"
!
"!

Figure 6.1: Subcase 2B: LM HM = (LM H)M

by Dedekinds Modular Law (Lemma 1.9). Now LM H is a -group (as a
subgroup of H) and
|(LM H)M |
|LM H|
|M |
=
= |M |
|LM H M |

|LM : LM H| =

(since H M = 1 as they have coprime order). Hence LM H is a Hall

-subgroup of LM .
If LM < G, we can apply induction to the group LM to see that some
conjugate of the Hall -subgroup LM H contains the -subgroup L:
L # (LM H)x # H x
for some x G (indeed we could pick x LM ). We would then be done.
So suppose LM = G. Then as L M = 1 (they have coprime order),
|G| = |LM | = |L| |M |,
so |L| = mq /q = m. Also, M # N , so G = LM = LN . Thus
|G| = |LN | =
so
|L N | =

|L| |N |
,
|L N |

m p q
|L| |N |
=
= p .
|G|
mq

Thus LN is a Sylow p-subgroup of N . By Sylows Theorem, it is conjugate

to the Sylow p-subgroup P which we already know about, say
L N = Px

where x N G.

Now L N ! L, since N $ G, so
L # NG (L N ) = NG (P x ) = NG (P )x = H x .

79

Thus L is contained in some conjugate of our Hall -subgroup H. This

completes the proof of (i) and (iii).
We have now shown that if G is a finite soluble group, then G has a
Hall -subgroup H, and every -subgroup of G is contained in a conjugate
of H. It remains to prove (ii). Let K be any Hall -subgroup of G. By (iii),
K # H x for some x G. But these subgroups have the same order, so we
deduce K = H x and so part (ii) of Theorem 6.34 holds.
This completes the proof of Halls Theorem.
!

6.2.3

Sylow systems and Sylow bases

We now examine some consequences of Halls Theorem. Specifically, we will

see how the Sylow subgroups of a soluble group can be arranged to have
special properties.
Definition 6.36 If p N is prime, write p" for the set of all primes not
equal to p. A Hall p" -subgroup of a finite group G is a p-complement.
Note that 2" then denotes the set of all odd primes.
Let G be a finite group, with |G| = pn m where p does not divide m.
Then a Hall p" -subgroup H has order m, while a Sylow p-subgroup P has
order pn . As they have coprime orders, we see H P = 1 and therefore
|HP | = |H| |P | = |G|,
so
G = HP,

H P = 1.

In this situation H and P are complements to one another (see Definition 5.9), although neither subgroup is necessarily normal.
Example 6.37 Consider D215 . A Hall 2" -subgroup has order 30/2 = 15,
so the subgroup #$ is a complement to the Sylow 2-subgroup #$ of D215 .
Let G be a finite soluble group and write
|G| = pn1 1 pn2 2 . . . pnk k
where p1 , p2 , . . . , pk are the distinct prime factors of |G|. By Halls Theorem,
G has a Hall p" -subgroup for each prime. Let Q1 , Q2 , . . . , Qk be Hall
p"i -subgroups for i = 1, 2, . . . , k, respectively. Then
|Qi | = |G|/pni i

and

80

|G : Qi | = pni i .

Claim: Q1 Q2 Qt is a Hall {pt+1 , . . . , pk }-subgroup of G.

Proof of Claim: This is certainly true when t = 1. Assume inductively
that the intersection H = Q1 Q2 Qt is a Hall {pt+1 , . . . , pk }-subgroup
of G. Then
n

t+1
|H| = pt+1
. . . pnk k

|G : H| = pn1 1 . . . pnt t .

and

Now apply Lemma 1.14: H and Qt+1 have coprime indices, so

n

t+1
|G : H Qt+1 | = |G : H| |G : Qt+1 | = pn1 1 . . . pnt t pt+1
.

t+2
Hence |H Qt+1 | = pt+2
. . . pnk k , so H Qt+1 = Q1 Qt+1 is a Hall
{pt+2 , . . . , pk }-subgroup of G. Thus the claim holds by induction.
!

In particular, Pk = Q1 Q2 Qk1 is a Hall {pk }-subgroup of G;

that is, a Sylow pk -subgroup of G. Generalising in the obvious way, we
deduce that
%
Pr =
Qi
i$=r

is a Sylow pr -subgroup of G (for r = 1, 2, . . . , k).

Now consider the two Sylow subgroups Pk1 and Pk . Firstly Pk1 Pk =
1 (since they have coprime orders), so
n

k1 nk
|Pk1 Pk | = |Pk1 | |Pk | = pk1
pk = |Pk Pk1 |.

Further, by construction, both Pk1 and Pk are contained in the intersection Q1 Q2 Qk2 and by our claim this intersection is a Hall
{pk1 , pk }-subgroup of G; that is,
n

k1 nk
|Q1 Q2 Qk2 | = pk1
pk .

Since it is a subgroup, this Hall subgroup is closed under products, so

Pk1 Pk , Pk Pk1 Q1 Q2 Qk2 .
Finally the subsets occurring in the previous inclusion all have the same
size, so we deduce
Pk1 Pk = Q1 Q2 Qk2 = Pk Pk1 .
Generalising in the obvious way, we deduce that for all r += s:
Pr Ps = Ps Pr .
Definition 6.38 Let G be a finite group and let p1 , p2 , . . . , pk be the
distinct prime factors of |G|.
81

(i) A Sylow system for G is a collection Q1 , Q2 , . . . , Qk such that Qi is a

Hall p"i -subgroup of G (for i = 1, 2, . . . , k).
(ii) A Sylow basis for G is a collection P1 , P2 , . . . , Pk such that Pi is a
Sylow pi -subgroup of G (for i = 1, 2, . . . , k) and such that
Pi Pj = Pj Pi

for all i and j.

We have shown:
Theorem 6.39 A finite soluble group possesses a Sylow system and a Sylow
basis.
!
Recall that the product HK of two subgroups is a subgroup if and only
if HK = KH. Consequently, if we start with a Sylow basis P1 , P2 , . . . , Pk
for a finite soluble group G, then we can form
Pi1 Pi2 . . . Pis
for any subset {i1 , i2 , . . . , is } {1, 2, . . . , k}. The fact that the Sylow subgroups in our Sylow basis permute ensures that this is a subgroup and it is
ni ni
n
easy to see that its order is pi1 1 pi2 2 . . . pisis . Thus we have formed a Hall
subgroup for the appropriate collection of primes. Hence a Sylow basis is a
nice collection of Sylow subgroups from which we may easily construct Hall
subgroups.
Philip Hall proved far more than these results. The final two theorems
of this section will not be proved: the first appears on Tutorial Sheet VI.
Theorem 6.40 (P. Hall) Let G be a finite soluble group. Then any two
Sylow bases for G are conjugate (that is, if P1 , P2 , . . . , Pk and R1 , R2 ,
. . . , Rk are two Sylow bases for G, where Pi and Ri are Sylow subgroups
for the same prime, then there exists x G such that Ri = Pix for all i).
This is much stronger than Sylows Theorem. The latter tells us that
each Ri is a conjugate of Pi . What the above theorem tells us is that when
the Sylow subgroups come from a Sylow basis then we can actually choose
the same element x to conjugate all the Sylow subgroups simultaneously.
Finally we have the following major converse to Halls Theorem.
Theorem 6.41 (P. Hall) Let G be a finite group which possesses a Hall
p" -subgroup for every prime p. Then G is soluble.
Putting Theorems 6.34 and 6.41 together, we see that a group is soluble
if and only if it has Hall -subgroups for all collections of primes. (In
particular, our observation that A5 was missing some Hall subgroups is no
longer surprising.)
82