Matrix groups

Peter J. Cameron

1 Matrix groups and group representations

These two topics are closely related. Here we consider some particular groups
which arise most naturally as matrix groups or quotients of them, and special
properties of matrix groups which are not shared by arbitrary groups. In repre-
sentation theory, we consider what we learn about a group by considering all its
homomorphisms to matrix groups.
This article falls roughly into two parts. In the first part we discuss proper-
ties of specific matrix groups, especially the general linear group (consisting of
all invertible matrices of given size over a given field) and the related classical
groups. In the second part, we consider what we learn about a group if we know
that it is a linear group.
Most group theoretic terminology is standard and can be found in any text-
book. A few terms we need are summarised in the next definition.
Let X and Y be group-theoretic properties. We say that a group G is locally
X if every finite subset of G is contained in a subgroup with property X; G is X-
by-Y if G has a normal subgroup N such that N has X and G/N has Y; and G is
poly-X if G has subgroups N0 = 1, N1 , . . . , Nr = G such that, for i = 0, . . . , r 1,
Ni is normal in Ni+1 and the quotient has X. Thus a group is locally finite if and
only if every finitely generated subgroup is finite; and a group is solvable if and
only if it is poly-abelian.

2 Introduction
This section contains some basic definitions about matrix groups.
A matrix group, or linear group, is a group G whose elements are invertible
n n matrices over a field F. The general linear group GL(n, F) is the group
consisting of all invertible n n matrices over the field F. So a group G is a matrix
group precisely when it is a subgroup of GL(n, F) for some natural number n and
field F. If V is a vector space of dimension n over F, we denote by GL(V ) the
group of invertible linear transformations of V ; thus GL(V ) = GL(n, F).

1
 Which groups are isomorphic to matrix groups? Auslander and Swan showed
that every polycyclic group is linear over Z; conversely, a solvable linear group
over Z is polycyclic.
Malcev showed that linearity of bounded degree is a local property:

Theorem 2.1 If every finitely generated subgroup of a group G is isomorphic to a

linear group of degree n, then G is isomorphic to a linear group of degree n. The
same assertion holds in any given characteristic.

Fact 2.2 Any free group is linear of degree 2 in every characteristic. More gen-
erally, a free product of linear groups is linear.

A matrix group G GL(V ) is said to be reducible if there is a G-invariant sub-

space U of V other than {0} and V , and is irreducible otherwise. A matrix group
G GL(V ) is said to be decomposable if V is the direct sum of two non-zero
G-invariant subspaces, and is indecomposable otherwise. If V can be expressed
as the direct sum of irreducible subspaces, then G is completely reducible. (Thus
an irreducible group is completely reducible.) If the matrix group G GL(n, F)
is irreducible regarded as a subgroup of GL(n, K) for any algebraic extension K
of F, we say that G is absolutely irreducible.

Theorem 2.3 (Maschkes Theorem) Let G be a finite linear group over F, and
suppose that the characteristic of F is either zero or coprime to |G|. If G is
reducible, then it is decomposable.

The image of a representation of a group is a matrix group. We apply de-

scriptions of the matrix group to the representation: thus, if : G GL(V ) is
a representation, and G is irreducible, indecomposable, absolutely irreducible,
etc., then we say that the representation is irreducible, etc.
Maschkes Theorem immediately extends to show that a locally finite group
in characteristic zero is completely reducible.

Theorem 2.4 (Cliffords Theorem) Let G be an irreducible linear group on a

vector space V of dimension n, and let N be a normal subgroup of G. Then
V is a direct sum of minimal N-spaces W1 , . . . ,Wd permuted transitively by G. In
particular, d divides n; the group N is completely reducible; and the linear groups
induced on Wi by N are all isomorphic.

It follows that a normal (or even a subnormal) subgroup of a completely re-

ducible linear group is completely reducible.
In general, a linear group G on V has a unipotent normal subgroup U such
that G/U is isomorphic to a completely reducible linear group on V ; the subgroup

2
U is nilpotent of class at most n 1, where n = dim(V ). (A matrix is unipotent
if all its eigenvalues are 1; a linear group is unipotent if all its elements are. A
unipotent linear group is conjugate (in the general linear group) to a group of
upper unitriangular matrices.)
A theorem of Malcev asserts that complete reducibility is a local property:
Theorem 2.5 If every finitely generated subgroup of the linear group G is com-
pletely reducible, then G is completely reducible.
Over an algebraically closed field, we have the following algebraic criterion
for irreducibility:
Theorem 2.6 Let G be a linear group on F n , where F is algebraically closed.
Then G is irreducible if and only if the elements of G span the space of all n n
matrices over F.

3 The general and special linear groups

In this section we consider the largest matrix group for given size of matrices and
given field, the general linear group, and its normal subgroup the special linear
group.
If F is the finite field of order q (the Galois field GF(q)), then we write
GL(n, q) for GL(n, F).
The special linear group SL(n, F) is the normal subgroup of GL(n, F) con-
sisting of matrices of determinant 1. The projective general linear group and
projective special linear group PGL(n, F) and PSL(n, F) are the quotients of
GL(n, F) and SL(n, F) by their normal subgroups Z and Z SL(n, F), respec-
tively, where Z is the group of non-zero scalar matrices. If F = GF(q), we write
SL(n, q) for SL(n, F), etc.
Theorem 3.1 Given any two ordered bases for the vector space V , there is a
unique element of GL(V ) carrying the first to the second.
From this theorem it follows that the order of GL(n, q) is equal to the number
of ordered bases of GF(q)n , namely
n1 n1
| GL(n, q)| = (qn qi ) = qn(n1)/2 (qni 1).
i=0 i=0

A transvection is a linear transformation t on V with all eigenvalues equal

to 1 and satisfying rank(t 1) = 1. A transvection has the form
t : x 7 x + (x f )v

3
where v V , f V , and v f = 0. Note that transvections have determinant 1, and
so lie in SL(V ).

Theorem 3.2 The group SL(n, F) is generated by transvections, for any n 2

and any field F.

Theorem 3.3 The group PSL(n, F) is simple for all n 2 and all fields F, except
for the two cases PSL(2, 2) and PSL(2, 3).

Fact 3.4 Basic facts from linear algebra about similarity of matrices can be in-
terpreted as statements about conjugacy classes in GL(n, F). For example, two
non-singular matrices are conjugate in GL(n, F) if and only if they have the same
invariant factors. If F is algebraically closed, then two non-singular matrices are
conjugate in GL(n, F) if and only if they have the same Jordan form.

4 The BN structure of the general linear group

The general linear groups GL(n, F), for n 2, contain a configuration of sub-
groups known as a BN-pair or Tits system. Our treatment is descriptive, but an
axiomatic approach is also possible.
Let G be the general linear group GL(n, F). Let B be the group of upper-
triangular matrices in G; U the group of strictly upper-triangular matrices (with
diagonal entries 1); H the group of diagonal matrices; and N the normaliser of H
in G, the group of monomial matrices (those having a unique non-zero element in
each row or column). The following properties are easily verified:

B = UH, B N = H;

N/H is isomorphic to the symmetric group Sn .

Let si be the transposition (i, i + 1) in Sn , for i = 1, . . . , n. We can take as coset

representative of si in N the reflection which interchanges the ith and (i+1)st basis
vectors; we denote this element also by si . The elements s1 , . . . , sn1 generate
Sn , and indeed Sn is a Coxeter group: a presentation is given by the relations
s2i = 1 (for i = 1, . . . , n 1), (si si+1 )3 = 1 (for i = 1, . . . , n 2); and (si s j )2 = 1 if
| j i| > 1.
Now it can be shown that G = BNB. Moreover, there are n! double cosets
BxB; we can take as double coset representatives the elements of Sn (regarded as
products of the generators si ). Furthermore, for any w Sn and i = 1, . . . , n 1,
we have
BwBsi B BwB Bwsi B,

4
from which it follows that hB, s1 , . . . , sn1 i = G.
The subgroups B and W = N/H are known as the Borel subgroup and Weyl
group of G. From the relation above, many strong properties can be deduced:
here is an example.
Theorem 4.1 Any subgroup of G containing B has the form
PI = hB, si : i Ii
for some subset I of {1, . . . , n 1}. Thus there are 2n1 such subgroups.
The subgroups PI are the parabolic subgroups of G (relative to the given BN-
pair). The maximal parabolic subgroups are those for which I = {1, . . . , n 1} \
{ j} for some j; it is easy to see that in this case PI is the stabiliser of the subspace
spanned by the last n j basis vectors.
More generally, with respect to any basis of V there is a BN-structure. We use
the terms Borel subgroup and parabolic subgroup to refer to the subgroups
defined with respect to an arbitrary basis. By Theorem 3.1, for example, all the
Borel subgroups of GL(V ) are conjugate. The maximal parabolic subgroups are
precisely the maximal reducible subgroups.

5 Classical groups
The classical groups form several important families of linear groups. We give
a brief description here, and refer to the books [7, 17] or the article [9] for more
details.
Let be an automorphism of F. A -sesquilinear form on the F-vector space
V is a function B : V V F which is linear as a function in the first variable
and -semilinear in the second. If is the identity, then B is a bilinear form. A
sesquilinear form B is non-degenerate if B(v, w) = 0 for all w V implies v = 0.
A sesquilinear form B is -Hermitian if B(v, w) = B(w, v) for all v, w V . A
bilinar form B is alternating if B(v, v) = 0 for all v V .
In general, the radical of B is the subspace V = {v V : B(v, w) = 0 for all w
V . In general one can define a left and right radical, but for alternating or Hermi-
tian forms the two radicals coincide.
A quadratic form on the F-vector space V is a function Q : V F satisfying
Q( v) = 2 Q(v) for all F, v V ;
the function B : V V F defined by
B(v, w) = Q(v + w) Q(v) = Q(w)
is bilinear.

5
The quadratic form Q is non-singular if Q(v) = 0 and B(v, w) = 0 for all w V
imply that v = 0; equivalently, the form Q is non-zero on all vectors in the radical
of B except the zero vector.

Fact 5.1 If the characteristic of F is not 2, then the form B is symmetric (that is, 1-
Hermitian), and Q can be recovered from B by the formula Q(v) = 21 B(v, v). If the
characteristic is 2, however, the form B is alternating and does not determine Q;
there are many quadratic forms associated with a given alternating form.

A classical group over V is the subgroup of GL(V ) consisting of transfor-

mations preserving either a non-degenerate Hermitian or alternating form or a
non-singular quadratic form. We distinguish three types of classical groups:

orthogonal group: preserving a non-singular quadratic form Q;

symplectic group: preserving a non-degenerate alternating bilinear form B;

unitary group: preserving a non-degenerate -Hermitian form B, with 6= 1.

We denote a classical subgroup of GL(V ) by O(V ), Sp(V ), or U(V ) depending on

type. If necessary, we add extra notation to specify which particular form is being
used. If V = F n , we also write O(n, F), Sp(n, F) or U(n, F).
A formed space will denote a vector space carrying a form of one of the three
classical types specified above.
The Witt index of a formed space V is the dimension of the largest subspace
on which the form is identically zero. We also speak of the Witt index of the
corresponding classical group.
Classification of classical groups up to conjugacy in GL(n, F) is equivalent
to classification of forms of the appropriate type up to the natural action of the
general linear group together with scalar multiplication. Often this is a very diffi-
cult problem: the next theorem gives a few cases where the classification is more
straightforward.

Theorem 5.2 (a) A non-degenerate alternating form on V = F n exists if and

only if n is even, and all such forms are equivalent. So there is a unique
conjugacy class of symplectic groups in GL(n, F) if n is even (with Witt
index n/2), and none if n is odd.

(b) Let F = GF(q). Then, up to conjugacy, GL(n, q) contains one conjugacy

class of unitary subgroups (with Witt index bn/2c), one class of orthogonal
subgroups if n is odd (with Witt index (n 1)/2), and two classes if n is even
(with Witt indices n/2 and n/2 1).

6
 (c) A non-singular quadratic form on Rn is determined up to the action of GL(n, R)
by its signature. Its Witt index is min{s,t}, where s and t are the numbers of
positive and negative eigenvalues. So there are bn/2c + 1 conjugacy classes
of orthogonal subgroups of GL(n, R), with Witt indices 0, 1, . . . , bn/2c.

The analogue for the classical groups of Theorem 3.1 is Witts Lemma. An
isometry between subspaces of a formed space is a linear transformation preserv-
ing the value of the form.

Theorem 5.3 (Witts Lemma) Suppose that U1 and U2 are subspaces of the formed
space V , and h : U1 U2 is an isometry. Then there is an isometry g of V which
extends h if and only if (U1 V )h = U2 V .
In particular, if V = 0, then any isometry between subspaces of V extends to
an isometry of V .

From Witts Lemma it is possible to write down formulae for the orders of the
classical groups over finite fields similar to the formula we gave for the general
linear group.
The analogues of Theorems 3.2 and 3.3 hold for the classical groups with
non-zero Witt index. However, the situation is more complicated. Any symplectic
transformation has determinant 1, so Sp(2r, F) SL(2r, F). Moreover, Sp(2r, F)
is generated by symplectic transvections (those preserving the alternating form)
for r 2, except for Sp(4, 2). Similarly, the special unitary group SU(n, F) (the
intersection of U(n, F) with SL(n, F)) with positive Witt index is generated by
unitary transvections (those preserving the Hermitian form), except for SU(3, 2).
There are no orthogonal transvections except in characteristic 2, and it is
necessary to use the more complicated long root elements instead.
Another complication with the orthogonal groups is that the group generated
by these elements is not always the intersection of the orthogonal group with the
special linear group, but may be smaller. In general it is denoted by (n, F),
with possibly some additional notation to specify which quadratic form is being
considered.
Then it can be shown that, if the Witt index is at least 3, then the quotient of
Sp(n, F), SU(n, F) or (n, F) by the group of scalar matrices it contains is simple.
Like the general linear groups, the classical groups contain BN-pairs (configu-
rations of subgroups satisfying conditions like those in the previous section). The
difference is that the Weyl group W = N/H is not the symmetric group, but one
of the other types of Coxeter group (finite groups generated by reflections).
Although our treatment of classical groups has been as far as possible inde-
pendent of fields, it is worth making two remarks here.

7
 The theory can be extended to classical groups over rings. This has impor-
tant connections with algebraic K-theory. The book [7] gives details.

On the other hand, for most of mathematics, the classical groups over the
real and complex numbers are the most important, and among these, the real
orthogonal and complex unitary groups preserving positive definite forms
take pride of place: see [24].

A representation of a group G (over the complex numbers) is said to be uni-

tary if its image is contained in the unitary group (that is, preserves a positive
definite Hermitian form). It is known that every representation of a finite group is
equivalent to a unitary representation.

6 Aschbachers Theorem
Aschbachers Theorem gives a description of the maximal subgroups of the clas-
sical groups over finite fields. We state the result here just for GL(n, q), and refer
to [1, 10] for the general case. The theorem is the analogue for linear groups of
the ONanScott Theorem for permutation groups.
We define eight classes of subgroups of GL(n, q) as follows:

C1 : Stabilisers of subspaces of V . (These are precisely the maximal parabolic

subgroups of GL(n, q).)

C2 : Stabilisers of direct sum decompositions of V into subspaces of the same

dimension.

C3 : Semilinear groups over extension fields of GF(q) of prime degree: that is,
the product of GL(m, qk ) and the Galois group of GF(qk ) over GF(q) inside
GL(mk, q), where k is prime.

C4 : Stabilisers of tensor product decompositions V = V1 V2 .

C5 : Linear groups over subfields of GF(q) of prime index.

C6 : Normalisers of r-groups of symplectic type, where r is a prime different

from p.

C7 : Stabilisers of tensor product decompositions V =

Nt
i=1 Vi , where the Vi all
have the same dimension.

C8 : Classical groups.

8
Here, for a prime number r, an r-group is said to have symplectic type if it is a
central product of a cyclic group and an extraspecial group. The normaliser of
this group involves a symplectic group (if r is odd) or an orthogonal group (if r is
even); we assume that the action of this group is irreducible. The definition of the
other classes should be obvious.

Theorem 6.1 (Aschbachers Theorem for GL(n, q)) Let H be a subgroup of GL(n, q),
not containing SL(n, q). Then either
(a) H is contained in a member of one of the classes C1 C8 ; or
(b) H is absolutely irreducible and H modulo scalars is almost simple.

Liebeck [12] proved the following addition, showing that the groups in case
2
(b) are very small (recall that the order of GL(n, q) is roughly qn ).

Theorem 6.2 In case (b) of Aschbachers Theorem, |G| q3n .

7 Finite matrix groups

According to Cayleys Theorem, every group is isomorphic to a permutation
group (a subgroup of a symmetric group). However, not every group is isomor-
phic to a matrix group (a subgroup of a general linear group). In the remainder
of the survey, we will give some results about the structure of matrix groups. The
information will be divided into several sections: properties of finite and finitely
generated matrix groups; periodic matrix groups; and solvable and nilpotent ma-
trix groups. We conclude with a brief look at finitary groups, a generalization of
matrix groups.
References on matrix groups (most of them fairly old) include Dixon [6],
Suprunenko [16], and Wehrfritz [22].
We begin by observing that every finite group is isomorphic to a matrix group
(we can embed it in a finite symmetric group and then take the representation by
permutation matrices). However, the class of matrix groups of fixed degree in
characteristic zero is restricted by a theorem of Jordan:

Theorem 7.1 There is a function f on the natural numbers such that, if G is a

finite linear group of degree n over a field of characteristic zero (or not dividing
|G|), then G has an abelian normal subgroup of index at most f (n).
2
The function f (n) can be taken to be (cn)n for some constant c.
The precise analogue for non-zero characteristic is false: the group GL(n, q)
2
has order roughly qn and any abelian normal subgroup is contained in the centre.

9
However, Brauer and Feit [4] showed that there is an abelian normal subgroup
whose index is bounded by a function of n and the order of a Sylow p-subgroup
of G, where p is the characteristic.
Burnside proved that any finite subgroup of GL(n, Q) is conjugate to a sub-
group of GL(n, Z). More generally, a finite subgroup of GL(n, F) is conjugate to
a subgroup of GL(n, R) if R is a principal ideal domain with field of fractions F.

8 Finitely generated matrix groups

Perhaps the most important general result on finitely generated matrix groups is
the Tits alternative (Tits [20]). This shows that in a certain sense a finitely gen-
erated matrix group is either tame or wild.
Theorem 8.1 A finitely generated linear group either is solvable-by-finite, or con-
tains a nonabelian free subgroup.
This theorem has been the starting point for similar results for a number of
other classes of groups, including linear groups over division rings, mapping class
groups, groups associated with low-dimensional manifolds, and groups that are
residually of bounded rank. An algorithmic form of Tits theorem has been given
by Beals [2].
Among the many consequences of the Tits alternative, we mention just two.
The first is the observation of Tits himself that the growth of a finitely generated
linear group (the growth rate for the number of group elements which are words
of length n in the generators) is either polynomial or exponential. (For general
groups, an example of Grigorchuk shows that growth strictly between polynomial
and exponential is possible.) The second is a striking theorem of Larsen and
Lubotzky [11]:
Theorem 8.2 Let G be a finitely generated residually finite linear group over an
algebraically closed field. Suppose that the number of normal subgroups of G
of index at most n is bounded by a polynomial in n. Then either G is solvable-
by-finite, or G involves a product of simple algebraic groups of type G2 , F4 or
E8 .
We conclude with a couple more special properties of finitely generated matrix
groups.
Theorem 8.3 Let G be a finitely generated linear group.
(Malcev) G is Hopfian, that is, not isomorphic to any proper quotient;
(Rabin) G has solvable word problem.

10
9 Periodic matrix groups
Burnside showed that a finitely generated linear group of finite exponent in char-
acteristic zero is finite. The result extends to arbitrary characteristic as follows:

Theorem 9.1 Let G be a linear group of degree n over F, and suppose that G has
3
exponent e. Then G has a normal subgroup N of index at most en such that

if char(F) = 0, then N = 1;

if char(F) = p > 0, then N is a p-group of nilpotency class at most n 1

consisting of unipotent elements. In particular, N = 1 if G is completely
reducible.

This shows that a completely reducible linear group of finite exponent is lo-
cally finite. But more is true, as was shown by Schur in characteristic zero and
Kaplansky in general:

Theorem 9.2 A periodic linear group is locally finite.

Using this result, Jordans theorem extends immediately from finite to periodic
linear groups in characteristic zero (or not dividing the characteristic).
Winter [25] showed that a completely reducible periodic linear group over an
algebraically closed field is conjugate to a linear group over the algebraic clo-
sure of the prime subfield. It follows immediately that any completely reducible
periodic linear group is countable.
Wehrfritz [21] showed that analogues of Sylows theorems hold for periodic
linear groups: that is, for any prime p, such a group has maximal p-subgroups,
and any two such subgroups are conjugate. Other results of P. Hall and Wielandt
on Sylow theory for finite groups have been extended to this class of groups.
The simple locally finite matrix groups were determined by Thomas [18, 19]
and Borovik [3]. The proof uses the Classification of Finite Simple Groups. The
groups of Lie type include the classical linear groups and some exceptional
families associated with the exceptional simple Lie algebras over C or with ex-
ceptional automorphisms of their Dynkin diagrams: see Carter [5] for discussion.

Theorem 9.3 A locally finite simple linear group is a group of Lie type.

11
10 Solvable and nilpotent matrix groups
Malcev showed that the analogue of Jordans theorem holds for completely re-
ducible solvable groups: such a group has a normal abelian subgroup whose index
is bounded by a function of the degree. If we delete the assumption of complete
reducibility, we can assert that there is a triangularisable subgroup of bounded
index, so that the group is nilpotent-by-abelian-by-finite.
Zassenhaus proved that a solvable linear group of degree n has derived length
bounded by a function of n. (The best possible bounds for , namely

5 log9 (n 1) + D (n) 5 log9 (n 2) + D + 32 ,

for n 66, where D = 17 2 (15 log 2)/(2 log 3), are due to Newman [14].) It
follows that a locally solvable linear group is solvable.
The next theorem summarises some results of Platonov and Wehrfritz.

Theorem 10.1 Let G be a finitely generated linear group. Then

a locally nilpotent subgroup of G is nilpotent;

the Frattini subgroup of G is nilpotent;

the Fitting subgroup of G is the set of all right Engel elements of G.

Suprunenko showed that the centre of an irreducible nilpotent linear group has
index bounded by a function of the degree and the nilpotency class.

11 Finitary groups
I conclude with a brief account of a generalization of matrix groups which was
first introduced by Dieudonne but has had a lot of attention recently. Let V be
an infinite-dimensional vector space over a field F. An invertible linear trans-
formation g on V is called finitary if the image of g 1 has finite dimension
(equivalently, the subspace of fixed points of g has finite codimension). A group
of linear transformations of V is called finitary if all its elements are finitary.
A highlight of the theory of finitary groups is a result of Hall [8], which ex-
tends Thomas result (Theorem 9.3) on linear groups.

Theorem 11.1 A simple locally finite finitary group is either an alternating group
or a finitary analogue of a classical matrix group.

12
 Wehrfritz has generalized many basic results on complete reducibility, Clifford
theory, etc., of linear groups to finitary groups. He has also extended the concept
of finitary groups from vector spaces to modules over rings. A rich and complex
theory emerges: see [23] for some results.
A survey of finitary groups appears in Phillips [15].

Acknowledgement I am grateful to Professor B. A. F. Wehrfritz for helpful

comments on this article.

References
[1] M. Aschbacher, On the maximal subgroups of the finite classical groups,
Invent. Math. 76 (1984), 469-514.

[2] R. Beals, Algorithms for matrix goups and the Tits alternative, J. Comput.
System Sci. 58 (1999), 260279.

[3] A. V. Borovik, Periodic linear groups of odd characteristic, Dokl. Akad. Nauk
SSSR 266 (1982), 12891291. (English translation: Soviet Math. Dokl. 26
(1982), 484486.)

[4] R. Brauer and W. Feit, An analogue of Jordans theorem in characteristic p,

Ann. Math. 84 (1966), 119131.

[5] R. W. Carter, Simple Groups of Lie Type, Wiley, New York, 1972.

[6] J. D. Dixon, The Structure of Linear Groups, Van Nostrand Reinhold, Lon-
don, 1971.

[7] A. Hahn and T. OMeara, The Classical Groups and K-Theory, Springer-
Verlag, Berlin, 1989.

[8] J. I. Hall, Locally finite simple groups of finitary linear transformations, in

Finite and locally finite groups (Istanbul, 1994), 147188, NATO Adv. Sci.
Inst. Ser. C Math. Phys. Sci., 471, Kluwer Acad. Publ., Dordrecht, 1995.

[9] L. Kramer, Buildings and classical groups, in Tits Buildings and the Model
Theory of Groups (ed. K. Tent), London Math. Soc. Lecture Notes 291, Cam-
bridge University Press, Cambridge, 2002.

[10] P. B. Kleidman and M. W. Liebeck, The Subgroup Structure of the Finite

Classical Groups, London Math. Soc. Lecture Notes 129, Cambridge Uni-
versity Press, Cambridge, 1990.

13
[11] M. Larsen and A. Lubotzky, Normal subgroup growth of linear groups: the
(G2 , F4 , E8 )-Theorem, Algebraic groups and arithmetic, 441468, Tata Inst.
Fund. Res., Mumbai, 2004.
[12] M. W. Liebeck, On the orders of maximal subgroups of the finite classical
groups, Proc. London Math. Soc. (3) 50 (1985), 426446.
[13] A. I. Malcev, On faithful representations of infinite groups of matrices, Mat.
Sb. 8 (1940), 405422; English translation Amer. Math. Soc. Transl. (2) 45
(1965), 118.
[14] M. F. Newman, The soluble length of soluble linear groups, Math Z. 126
(1972), 5970.
[15] R. E. Phillips, Finitary linear groups: a survey, in Finite and locally finite
groups (Istanbul, 1994), 111146, NATO Adv. Sci. Inst. Ser. C Math. Phys.
Sci., 471, Kluwer Acad. Publ., Dordrecht, 1995.
[16] D. Suprunenko, Soluble and nilpotent linear groups (transl. K. A. Hirsch),
Amer. Math. Soc., Providence, 1963.
[17] D. E. Taylor, The Geometry of the Classical Groups, Heldermann Verlag,
Berlin, 1992.
[18] S. Thomas, An identification theorem for the locally finite nontwisted
Chevalley groups, Arch. Math. (Basel) 40 (1983), 21131.
[19] S. Thomas, The classification of the simple periodic linear groups, Arch.
Math. (Basel) 41 (1983), 103116.
[20] J. Tits, Free subgroups in linear groups, J. Algebra 20 (1972), 250270.
[21] B. A. F. Wehrfritz, Sylow theorems for periodic linear groups. Proc. London
Math. Soc. (3) 18 (1968) 125140.
[22] B. A. F. Wehrfritz, Infinite Linear Groups, Queen Mary College Maths
Notes, London, 1969.
[23] B. A. F. Wehrfritz, Artinian-finitary groups over commutative rings and non-
commutative rings, J. London Math. Soc. (2) 70 (2004), 325340.
[24] H. Weyl, The Classical Groups, Princeton University Press, Princeton, 1939
(reprint 1997).
[25] D. J. Winter, Representations of locally finite groups, Bull. Amer. Math. Soc.
74 (1968), 145148.

14