The proof is from the book Groups and characters by Larry C. Grove.

The simplicity of PSL(n, q)

If V is n-dimensional vector space over a eld F then GL(V ) denotes the general linear group of
V . Choosing a basis for V provides an isomorphism of GL(V ) with the group GL(n, F) of all
invertible n n matrices over F.
The determinant is a homomorphism from GL(V ) onto the multiplicative group F

; its kernel is
the subgroup SL(V ), the special linear group.
The center of GL(V ) is Z(GL(V )) = a1 : a F

and the center of SL(V ) is Z(SL(V )) = a1 :

a F

, a
n
= 1.
If F is nite, with q elements, the the matrix groups are denoted by GL(n, q) and SL(n, q).
If 0 = v V write [v] for the line Fv = av : a F through the origin spanned by v, and call
it a projective point. The set of all distinct projective points [v] is called the projective space of
dimension n 1 based on V , and is denoted by P
n1
(V ), or P(V ). There is a natural action of
GL(V ) on P(V ), given by [v] = [v] for all GL(V ), [v] P(V ). The kernel of the action is
Z(GL(V )), and likewise that the kernel of the action of SL(V ) on P(V ) is Z(SL(V )).
Dene the projective special linear group on V to be
PSL(V ) = SL(V )/Z(SL(V )).
This group acts faithfully on P(V ).
Our goal is to show that except for PSL(2, 2) and PSL(2, 3) every PSL(V ) is a simple group.
To accomplish this well use Iwasawas theorem. But before we formulate and prove this theorem,
the denition of primitive group action is needed.
Let a group G act on a set S. For x G and s S we will write s
x
to denote the result of the
action of x on s. With that convention we have s
1
= s and s
xy
= (s
x
)
y
. Dene a block to be a
subset B S such that B = S, [B[ > 1, and if x G then either B
x
= B or B B
x
= . If G is
transitive on S and has no blocks we say that G is primitive on S.
Theorem (Iwasawa) Suppose that G is faithful and primitive on S and G
(1)
= G. Fix s S and
set H = Stab
G
(s). Suppose there is a solvable subgroup K H such that G = 'K
x
: x G`.
Then G is simple.
Proof. Suppose that 1 = N G. We will show that in this case N = G and therefore G has no
proper normal subgroups.
Step 1. N is transitive on S.
Since G is faithful on S and N = 1 there is an N-orbit in S having more than one element, say
B = Orb
N
(a) and [B[ 2. If x G then B
x
= a
Nx
= a
xN
(since N G). Thus B
x
= Orb
N
(a
x
).
Since S is partitioned into its N-orbits we have either B
x
= B or B
x
B = . But G is primitive
on S hence we must have B = S, or else B would be a block for G. Therefore Orb
N
(a) = B = S
and this means that N is transitive on S.
Step 2. G = Stab
G
(s)N for all s S. In particular, G = HN.
If x G then s
x
= s
y
for some y N since N is transitive on S. Thus s
xy
1
= s, so xy
1

Stab
G
(s) and x Stab
G
(s)y Stab
G
(s)N. Hence G = Stab
G
(s)N.
Step 3. G = KN.
G = HN = NH (since N G), and G = 'K
x
: x NH`. K H, hence for n N and h H
we have K
nh
= nhKh
1
n
1
=

K
h

n
= K
n
. Thus G = 'K
n
: n N`. Every g G can
be written in the form g = k
n
1
1
k
n
2
2
. . . k
n
r
r
= n
1
k
1
n
1
1
n
2
k
2
n
1
2
. . . n
r
k
r
n
1
r
. Since N G we have
partial commutativity kn = nk hence g = kn for some k K and n N. Thus G = KN.
1
Step 4. N = G.
Since K is solvable K
(m)
= 1 for some m.
(KN)
(1)
=

k
1
n
1
k
2
n
2
n
1
1
k
1
1
n
1
2
k
1
2
: k
1
, k
2
K, n
1
, n
2
N

=

k
1
k
2
k
1
1
k
1
2
n : k
1
, k
2
K, n N

K
(1)
N.
Using induction we obtain that (KN)
(l)
K
(l)
N for all l. Since G = G
(1)
we have G = G
(m)
=
(KN)
(m)
K
(m)
N = N. Therefore N = G and this ends the proof.
In order to apply Iwasawas theorem to PSL(V ) we must show that PSL(V ) is primitive on P(V ),
PSL(V )
(1)
= PSL(V ), and there exists a normal solvable subgroup in Stab
PSL(V )
([v]) whose conju-
gates in PSL(V ) generate PSL(V ). To achieve this goal we will introduce transvections and prove
some results for them.
A hyperplane in V is any subset of codimension 1. If 1 = GL(V ) then is called a transvection
if there is a hyperplane W such that [
W
= 1
W
and v v W for all v V ; W is called the
xed hyperplane of .
If is a transvection with xed hyperplane W choose a basis for V consisting of rst some v
1

V ` W and then a basis v
2
, . . . , v
n
for W. It is clear from the matrix representing relative to
this basis that det = 1, so SL(V ).
Proposition 1 The inverse of a transvection is a transvection. Suppose that V is a subspace of a
space V
1
, that v V
1
` V , and that is a transvection on V with xed hyperplane W, then can
be extended to a transvection
1
on V
1
whose xed hyperplane W
1
contains v.
Proof. If is a transvection then w = w for all w W. Multiplying both sides of this equality
by
1
= 1 we obtain w =
1
w for all w W. Also we have vv = w W, hence v
1
v = w,
and we obtain
1
v v = w W. Thus
1
is a transvection.
Now prove the second part of the proposition. Let v
1
, v
2
, . . . , v
k
be a basis for V such that
v
2
, . . . , v
k
is a basis for W. We can choose a basis for V
1
consisting of the following vectors
v
1
, v
2
, . . . , v
k
, v
k+1
= v, . . . , v
n
. Consider a hyperplane W
1
spanned by v
2
, . . . , v
n
and extend
on V
1
by setting
1
[
V
= [
V
,
1
v
i
= v
i
for all i > k. Then
1
[
W
1
= 1
W
1
and
1
v
1
v
1
= v
1
v
1

W W
1
. So
1
is a transvection with xed hyperplane containing v
k+1
= v.
Proposition 2 If u and v are linearly independent in V then there is a transvection with u = v.
Proof. Choose a hyperplane W in V with u v W but u / W, and dene by means of
[
W
= 1
W
, u = v. If x V write x = au + w, where a F and w W. Then x x =
av + w au w = a(v u) W, so is a transvection.
Proposition 3 Suppose that W
1
and W
2
are two distinct hyperplanes in V and that v V `
(W
1
W
2
). Then there is a transvection with W
1
= W
2
and v = v.
Proof. Note that W
1
+ W
2
= V , so dimW
1
W
2
= dimW
1
+ dimW
2
dim(W
1
+ W
2
) = n 2
and W = W
1
W
2
+ Fv is another hyperplane. Write v = x + y, with x W
1
and y W
2
. Then
x / W
2
, so W
1
= W
1
W
2
+Fx, and likewise W
2
= W
1
W
2
+Fy. Thus V = W
1
W
2
+Fx+Fy.
It follows that x / W, or else y = v x is also in W and hence V W, a contradiction. Dene
via [
W
= 1
W
and x = y. If z V write z = ax + w, where a F and w W. Then
z z = ay +w ax w = a(y +x) = av W. So is a transvection, v = v since v W,
and W
1
= (W
1
W
2
+ Fx) = W
1
W
2
Fy = W
2
.
Theorem 1 The set of transvections generates SL(V ).
2
Proof. Fix SL(V ), then choose a hyperplane W in V and choose v V ` W. If v and
v are linearly independent then by Proposition 2 there is a transvection
1
with
1
v = v. If v
and v are linearly dependent then rst choose a transvection
0
so that v and
0
v are linearly
independent, then a transvection

1
so that

0
v = v, and set
1
=

0
. Thus in either case we
have
1
v = v and
1
is a product of transvections. Note that v /
1
W (since v V ` W and

1
v = v). If
1
W = W set
2
= 1
V
. If
1
W = W apply Proposition 3 to get a transvection

2
with
2

1
W = W and
2
v = v. Set =
2

1
. Since SL(V ), W = W, and v = v it
follows that [
W
SL(W). Now use induction on n = dimV . If n = 2 then [
W
= 1
W
(since
SL(W) = 1
W
when dimW = 1), so = 1 and =
1
1

1
2
. Proposition 1 implies that is a
product of transvections. If n > 2 then by induction [
W
is a product of transvections on W, each
of which extends to a transvection xing v on V by Proposition 1. Thus is the product of the
extended transvections, and =
1
1

1
2
, a product of transvections.
Proposition 4 If
1
and
2
are transvections on V then they are conjugate in GL(V ). If n > 2
they are conjugate in SL(V ).
Proof. For i = 1, 2 write W
i
for the xed hyperplane of
i
, choose x
i
V ` W
i
, and set
w
i
=
i
x
i
x
i
W
i
. Choose bases w
1
, u
3
, . . . , u
n
for W
1
and w
2
, v
3
, . . . , v
n
for W
2
. For each
a F

dene
a
GL(V ) by setting
a
x
1
= x
2
,
a
w
1
= w
2
,
a
u
i
= v
i
for 3 i n 1,
and, if n > 2,
a
u
n
= av
n
. Then
a

1
a
(x
2
) =
a

1
(x
1
) =
a
(x
1
+ w
1
) = x
2
+ w
2
=
2
(x
2
),

1
a
(w
2
) =
a

1
(w
1
) =
a
w
1
= w
2
=
2
(w
2
), and
a

1
a
(v
i
) =
a

1
(u
i
) =
a
u
i
= v
i
=

2
(v
i
). So
a

1
a
and
2
agree on the basis x
2
, w
2
, v
3
, . . . , v
n
, and hence
a

1
a
=
2
.
If n > 2 we may set b = (det
1
)
1
and obtain
b
SL(V ).
Proposition 5 Suppose that dimV = 2, and let v
1
, v
2
be any basis for V . Every transvection
is conjugate in SL(V ) to one whose matrix relative to v
1
, v
2
is of the form

1 0
a 1

, a F

.
Proof. If is a transvection with xed hyperplane W choose v V ` W and set w = v v W.
Relative to the basis v, w the matrix representing is

1 0
1 1

. If M is the matrix that

represents relative to v
1
, v
2
then there is a matrix B in GL(2, F) with B
1
MB =

1 0
1 1

.
If det B = a
1
we may set A =

a 0
0 1

. Then BA SL(2, F), and

(BA)
1
M(BA) = A
1

1 0
1 1

A =

1 0
a 1

.
Theorem 2 If n 3 and G = SL(V ) then G
(1)
= G, and also if G = PSL(V ) then G
(1)
= G.
Proof. If we exhibit a transvection in G
(1)
then all will be done since by Proposition 4 all
transvections are conjugate in G, hence G
(1)
will contain all transvections, and by Theorem 1
transvections generate SL(V ).
Choose a basis v
1
, . . . , v
n
for V and dene
1
,
2
via

1
: v
1
v
1
v
2
, v
i
v
i
if 2 i n,

2
: v
1
v
1
, v
2
v
2
v
3
, v
i
v
i
if 3 i n.
Then

1
1

1
2
: v
1
v
1
v
3
, v
i
v
i
if 2 i n,
so
1

1
1

1
2
is a transvection in G
(1)
.
3
PSL(V ) = SL(V )/Z(SL(V )), hence
PSL(V )
(1)
=
SL(V )
(1)
Z(SL(V ))
Z(SL(V ))
=
SL(V )
Z(SL(V ))
= PSL(V )
.
Theorem 3 If n = 2, [F[ > 3, and G = SL(V ), then G
(1)
= G.
Proof. Choose a basis v
1
, v
2
for V and choose a F

, a = 1. Dene SL(V ) via

(v
1
) = a
1
v
1
, (v
2
) = av
2
, and for each b F

dene
b
SL(V ) via
b
(v
1
) = v
1
+ bv
2
,

b
(v
2
) = v
2
. Then
b

1
b
is represented by the matrix

a
1
0
0 a

1 0
b 1

a 0
0 a
1

1 0
b 1

1 0
ba(a a
1
) 1

.
The theorem follows from Theorem 1 and Proposition 5, since b F

is arbitrary.
Proposition 6 If n 2 then SL(V ) is primitive on the projective space P(V ).
Proof. Take [v
1
] = [v
2
] and [w
1
] = [w
2
] in P(V ), so v
1
, v
2
and w
1
, w
2
are linearly independent
sets in V . If n = 2 they are bases. If n 3 set V
1
= Fv
1
+Fv
2
and V
2
= Fw
1
+Fw
2
. Then either
V
1
= V
2
= V or else V
1
= V
2
, in which case V
1
V
2
is not a subspace in either case V
1
V
2
= V . If
v
3
V ` (V
1
V
2
) then both v
1
, v
2
, v
3
and w
1
, w
2
, w
3
are linearly independent. The argument
may be repeated to obtain two bases for V , v
1
, . . . , v
n
and w
1
, . . . , w
n
. For any b F

dene

b
GL(V ) via v
1
bw
1
, v
2
w
2
, v
i
v
i
for 3 i n. If w
j
=

i
a
ij
v
i
, j = 1, 2, then
det
b
= b(a
11
a
22
a
12
a
21
). Choose b so that det
b
= 1; then
b
SL(V ), and it carries [v
1
] to
[w
1
], [v
2
] to [w
2
].
Suppose B P(V ), with [B[ > 1 and B = P(V ). Choose [v
1
], [v
2
] B and [w] P(V ) ` B. Then
choose SL(V ) with ([v
1
]) = [v
1
] and ([v
2
]) = [w]. Then [v
1
] (B) B so (B) B = ,
and [w] (B) ` B, so (B) = B. Thus B is not a block for SL(V ), hence SL(V ) has no blocks
and it is primitive.
Proposition 7 If 0 = v V and A = Stab
SL(V )
([v]) then A has an abelian normal subgroup B
whose conjugates in SL(V ) generate SL(V ).
Proof. Choose a hyperplane W with v / W, so V = W + Fv. If A and w W write
w =

w + a
w
v, with

w W and a
w
F. Let x = aw
1
+ bw
2
then (aw
1
+ bw
2
) =

(aw
1
+ bw
2
) + a
x
v and (aw
1
+ bw
2
) = a(w
1
) + b(w
2
) = a

w
1
+ b

w
2
+ (aa
w
1
+ ba
w
2
)v.
Therefore

(aw
1
+ bw
2
) = a

w
1
+ b

w
2
since v / W. Also we have v =

v + a
v
v, on the
other hand v = a
v
v since A = Stab
SL(V )
([v]), so

v = 0.
1
w = (
1
)

w + b
w
v, hence
w =
1
(w) = (
1
)

(w) = (
1
)

w + a
w
v) = (
1
)

w. Analogously

(
1
)

w = w,
therefore

GL(W). Consider a map :

. (
1

2
)w = (
1

2
)

w + a
w
v, (
1

2
)w =

1
(
2
w) =

1
(

2
w+b
w
v) =

1
(

2
w)+a

2
w
v, hence (
1

2
)

2
, therefore is a homomorphism
form A into GL(W). Thus B = ker is a normal subgroup of A.
Choose a basis w
1
, . . . , w
n1
for W. For any b F

dene
b
A via v
1
w
1
+ bv, w
i

w
i
for 2 i n 1, and v v. Then
b
is a transvection (with hyperplane spanned by
w
2
, . . . , w
n1
, v), and if w W then
b
w = w + bv, so

b
= 1
W
and
b
B. If n = 2 the matrix
representing
b
relative to the basis w
1
, v is

1 0
b 1

, and b F

is arbitrary. It follows from

Theorem 1 and Propositions 4 and 5 that the conjugates of B generate SL(V ) for all n 2. If
B then

= 1
W
, so w
i
= w
i
+ a
i
v. Thus, if
1
,
2
B, their representing matrices relative
to the basis w
1
, . . . , w
n1
, v have the partitioned form

I 0
u
1
1

and

I 0
u
2
1

, where I is the
(n 1) (n 1) identity matrix and u
1
,u
2
are 1 (n 1) matrices. Since

I 0
u
1
1

I 0
u
2
1

I 0
u
1
+ u
2
1

I 0
u
2
1

I 0
u
1
1

4
we see that B is abelian.
Theorem 4 Except for PSL(2, 2) and PSL(2, 3), every PSL(V ) is a simple group.
Proof. Choose v = 0 in V , and take A = Stab
SL(V )
([v]) and B A as in Proposition 7. Write
Z = Z(SL(V )) and set H = A/Z = Stab
PSL(V )
([v]), K = BZ/Z H. We have that PSL(V )
acts faithfully on P(V ). It is also primitive on P(V ) since by Proposition 6 SL(V ) is primitive
on P(V ). K H = Stab
PSL(V )
([v]) and K is abelian, hence it is solvable. The conjugates of K
generate PSL(V ). Then all the conditions of Iwasawas Theorem are met, and PSL(V ) is simple.
5