Chapter 4

The Jordan–Hölder Theorem

Definition 4.1 Let G be a group. A subnormal series of G is a finite chain

of subgroups
G = G0 > G1 > G2 > · · · > Gn = 1
such that Gi+1 is a normal subgroup of Gi for i = 0, 1, . . . , n − 1. The
collection of quotient groups

G0 /G1 , G1 /G2 , . . . , Gn−1 /Gn

are the factors of the series, and the length of the series is n.
Note that we do not require each subgroup in the subnormal series to be
normal in the whole group, only that it is normal in the previous subgroup
in the chain. A normal series is a series where Gi is a normal subgroup of G
for all i. Note also that the length n is also the number of factors occurring.
We will be interested in three different types of subnormal series in this
course, and for all three we will require special properties of the factors. The
first case is where the factors are all required to be simple groups.

Definition 4.2 A composition series for a group G is a finite chain of sub-

groups
G = G0 > G1 > · · · > Gn = 1
such that, for i = 0, 1, . . . , n − 1, Gi+1 is a normal subgroup of Gi and the
quotient group Gi /Gi+1 is simple. The quotient groups

G0 /G1 , G1 /G2 , . . . , Gn−1 /Gn

are called the composition factors of G.

Example 4.3 Let G = S4 , and consider the following chain of subgroups:

S4 > A4 > V4 > "(1 2)(3 4)# > 1. (4.1)

44
We know that A4 ! S4 and V4 ! A4 . Since V4 is an abelian group,
"(1 2)(3 4)# ! V4 . Certainly 1 ! "(1 2)(3 4)#. Hence (4.1) is a series of
subgroups, each normal in the previous one. We can calculate the order of
each subgroup, and hence calculate the order of the quotient groups:

|S4 /A4 | = 2
|A4 /V4 | = 3
|V4 /"(1 2)(3 4)#| = 2
|"(1 2)(3 4)#| = 2.

Thus the quotients are all of prime order. We now make use of the fact that
a group G of prime order p is both cyclic and simple (see Example 3.6), to
see that the factors for the series (4.1) are cyclic simple groups. Thus (4.1)
is composition series for S4 , with composition factors

C2 , C3 , C2 , C2 .

Example 4.4 The infinite cyclic group G has no composition series.

Proof: Let G = "x#, where |x| = ∞, and suppose

G = G0 > G1 > G2 > · · · > Gn = 1

is a composition series for G. Then Gn−1 is a non-trivial subgroup of G,

so Gn−1 = "xk # for some positive integer k (see Tutorial Sheet I). Hence
Gn−1 is an infinite cyclic group, so is not simple. This contradicts our series
being a composition series. !

Suppose that there are subgroups

G"N >M "1

with M ! N . (The idea here is that M and N will be successive terms in a

series which we are testing to see whether or not it is a composition series.)
The Correspondence Theorem tells us that subgroups of N/M correspond to
subgroups of N which contain M . Furthermore, under this correspondence,
normal subgroups of N/M correspond to normal subgroups of N which
contain M . We conclude that

N/M is simple if and only if the only normal subgroups of N

containing M are N and M themselves.

Proposition 4.5 Every finite group possesses a composition series.

45
Proof: Let G be a finite group. We know at least one subnormal series,
namely G " 1. Let

G = G0 > G1 > · · · > Gn = 1 (4.2)

be a longest possible subnormal series of G. This certainly exists since G has

only finitely many subgroups, so only finitely many subnormal series. We
claim that (4.2) is a composition series.
Suppose that it is not. Then one of the factors, say Gj /Gj+1 , is not
simple. This quotient then possesses a non-trivial proper normal subgroup
and this corresponds to a subgroup N of G with

Gj+1 < N # Gj .

Then
G = G0 > G1 > · · · > Gj > N > Gj+1 > · · · > Gn = 1
is a subnormal series in G (note that Gj+1 ! N since Gj+1 ! Gj ) which is
longer than (4.2). This contradicts our assumption that (4.2) is the longest
such series. Hence (4.2) is indeed a composition series for G. !

On the other hand, we have an example of an infinite group (namely the

infinite cyclic group) which does not possess a composition series. There are
infinite groups that possess composition series, but infinite simple groups
(which necessarily occur as some of the composition factors) are much less
well understood than finite simple groups.
The important thing about composition series is that the composition
factors occurring are essentially unique. This is the content of the following
important theorem.

Theorem 4.6 (Jordan–Hölder Theorem) Let G be a group and let

G = G0 > G1 > G2 > · · · > Gn = 1

and
G = H0 > H1 > H2 > · · · > Hm = 1
be composition series for G. Then n = m and there is a one-one correspon-
dence between the two sets of composition factors

{G0 /G1 , G1 /G2 , . . . , Gn−1 /Gn }

and
{H0 /H1 , H1 /H2 , . . . , Hm−1 /Hm }
such that corresponding factors are isomorphic.

Proof: See Tutorial Sheet IV. !

46
 Hence, once we have determined one composition series for a (say, finite)
group, then we have uniquely determined composition factors which can be
thought of as the ways of breaking our original group down into simple
groups. This is analogous to a ‘prime factorisation’ for groups.
To recognise when we have a composition series, we need to be able
to recognise simple groups. We have already observed (Example 3.6) that
cyclic groups of prime order are simple. It is not hard to show that these
are all the abelian simple groups. The following was proved in MT4003.

Theorem 4.7 Let n " 5. Then the alternating group An is simple.

It is worth pointing out that much, much more is known. A mammoth
effort by hundreds of mathematicians from the 1950s to the 1980s succeeded
in classifying the finite simple groups. The complete proof runs to tens of
thousands of pages of extremely complicated mathematics. More work is
still being done to check, clarify and simplify the proof. Nevertheless, it is
generally accepted that this Classification is correct, though when relying
upon it a mathematician would normally state that he or she is doing so.

Theorem 4.8 (Classification of Finite Simple Groups) Let G be a fi-

nite simple group. Then G is one of the following:

(i) a cyclic group of prime order;

(ii) an alternating group An where n " 5;

(iii) one of sixteen infinite families of groups of Lie type;

(iv) one of twenty-six sporadic simple groups.

The groups of Lie type are essentially ‘matrix-like’ groups which preserve
geometric structures on vector spaces over finite fields. For example, the first
(and most easily described) family is

SLn (q)
PSLn (q) = .
Z(SLn (q))

To make it, first construct the group GLn (q) of invertible n × n matrices
with entries from the finite field F of order q. Next, take the subgroup of
matrices of determinant 1:

SLn (q) = { A ∈ GLn (q) | det A = 1 },

the special linear group. Now, factor by the centre of SLn (q), which consists
of all scalar matrices of determinant 1:

Z(SLn (q)) = { λI | λ ∈ F , λn = 1 }.

47
The result is PSLn (q). It can be shown that PSLn (q) is simple if and only if
either n = 2 and q " 4, or n " 3. The construction of the other 15 families
of groups of Lie type is similar but harder.
The twenty-six sporadic groups are:

Name Symbolic name Order

Mathieu M11 7 920
Mathieu M12 95 040
Janko J1 175 560
Mathieu M22 443 520
Janko J2 604 800
Mathieu M23 10 200 960
Higman–Sims HS 44 352 000
Janko J3 50 232 960
Mathieu M24 244 823 040
McLaughlin McL 898 128 000
Held He 4 030 387 200
Rudvalis Ru 145 926 144 000
Suzuki Suz 448 345 497 600
O’Nan O’N 460 815 505 920
Conway Co3 495 766 656 000
Conway Co2 42 305 421 312 000
Fischer Fi22 64 561 751 654 400
Harada–Norton HN 273 030 912 000 000
Lyons Ly 51 765 179 004 000 000
Thompson Th 90 745 943 887 872 000
Fischer Fi23 4 089 470 473 293 004 800
Conway Co1 4 157 776 806 543 360 000
Janko J4 86 775 571 046 077 562 880
Fischer Fi"24 1 255 205 709 190 661 721 292 800
Baby Monster B 4 154 781 481 226 426 191 177 580 544 000 000
Monster M see below

|M| = 808 017 424 794 512 875 886 459 904 961 710 757 005 754 368 000 000 000
Unsurprisingly, we omit the proof of Theorem 4.8.

We finish this chapter with a few examples.

Example 4.9 Let G be a finite abelian group of order n. Write

n = pr11 pr22 . . . prss
where p1 , p2 , . . . , ps are the distinct prime factors of n. If
G = G0 > G1 > G2 > · · · > Gm = 1

48
is a composition series, then the composition factors

G0 /G1 , G1 /G2 , . . . , Gm−1 /Gm

are abelian simple groups. They are therefore cyclic of prime order. Now

|G| = |G0 /G1 | · |G1 /G2 | · . . . · |Gm−1 /Gm |.

This must be the prime factorisation of |G| = n, and hence the composition
factors of G are

Cp1 , Cp1 , . . . , Cp1 , Cp2 , Cp2 , . . . , Cp2 , . . . , Cps , Cps , . . . , Cps .

! "# $ ! "# $ ! "# $
r1 times r2 times rs times

Although the Jordan–Hölder Theorem tells us that the composition fac-

tors are essentially uniquely determined, the composition series need not be
unique. For example, if G = "x# is cyclic of order 30, then there are several
different composition series; e.g.,

G = "x# > "x2 # > "x6 # > 1

where the composition factors are C2 , C3 and C5 , and

G = "x# > "x3 # > "x15 # > 1

where the composition factors are C3 , C5 and C2 ; etc.

Later in the course we shall characterise the finite groups whose compo-
sition factors are cyclic as being the soluble groups.
Our final example has a unique composition series:

Example 4.10 Let n " 5 and consider the symmetric group Sn of degree n.
We already know the following series:

Sn > An > 1 (4.3)

which has factors C2 and An . Both of these are simple groups, so (4.3) is
a composition series for Sn . It can be shown that Sn has precisely three
normal subgroups (namely those occurring in the above series) and hence
(4.3) is the only composition series for Sn .
The Jordan–Hölder Theorem again raises the question of how we put the
composition factors back together. We have a unique decomposition, but
how complicated is the reverse process? The answer turns out to be rather
difficult, but in the next chapter we shall meet some ways of creating new
groups and this will give some ways of putting the composition factors back
together.

49