The general linear group is a useful entry point into the mathematical realm of group theory (roughly, the abstract study of symmetries), since it ties that subject into the more familiar (for most undergraduates) linear algebra. Translating abstract notions of groups into more recognisable properties of matrices (or, in turn, linear maps) can make it easier to grasp them at first, and provides some indication of the power of such abstraction once it is subsequently applied to other groups (such as those of permutations).
There isn't one general linear group; rather a given example specifies two parameters- the general linear group GL(n,F) ( also written GLn(F) ) comprising the set of all square matrices M ∈ Matn,n(F) with non-zero determinant. For example, GL(2,R) is the set of all matrices
a b
c d
Where a,b,c,d are
real and ad-bc
≠ zero.
In other words, the general linear group contains only invertible matrices (since a matrix is invertible if and only if it has non-zero determinant). This restriction is necessary since if we simply take the set of all nXn matrices, the group axioms do not always hold. With this construction, and the familiar matrix multiplication as the group operation, these axioms can be easily verified:
For A,B,C ∈ GL(n,F) using matrix multiplication ∗
- Associativity: By definition of matrix multiplication, (A∗B)∗C = A∗(B∗C).
- Identity- ∃e∈GL(n,F) such that A∗e=A=e∗A: This relation is satisfied by the Identity matrix In. We can chose this matrix as e since it is of size nXn, has determinant 1 and (as any field contains 0 and 1) is therefore in GL(n,F).
- Inverses: for any A∈ GL(n,F), there is A-1 ∈Matn,n(F) with AA-1=I=A-1A, else A wouldn't be invertible and hence would have zero determinant (contradicts assumption of A being in the general linear group). We have already shown that I is our identity element e for the group; so all we require is that A-1 also be in GL(n,F). But if A is nXn in size, so is A-1; and as det(A) does not equal zero, det(A-1)=1/det(A) by properties of determinant is also non zero. We are done.
Some formulations of the group axioms require that closure be checked- that is, if A,B∈G for some group G, then A∗B should also be an element of G. Since our chosen operator, matrix multiplication, is defined on a larger set, that of all matrices, it would be wise to check this condition here to make sure that applying the operation doesn't cause us to 'fall out' of the group. If A and B ∈GL(n,F) then A∗B=AB is also an nXn matrix. By the product rule for determinants, det(A∗B)=det(AB)=det(A)det(B). This cannot be zero for det(A), det(B) non-zero, so AB satisfies the conditions for membership of GL(n,F).
The notion of the general linear group, besides providing a compact way of refering to the invertible matrices, is a starting point for several relations and groups of interest in linear algebra/group theory. For instance, if V is a vector space over F, the group GL(V) = {α | α:V→V is an invertible linear transformation} is isomorphic to GL(n,F).
If the field F is itself finite, then GL(n,F) will be finite for a given n. For instance, with F2={0,1} the group GL(2,F2) comprises the matrices
1 0 0 1 1 1 1 0 0 1 1 1
0 1 1 0 0 1 1 1 1 1 1 0
The determinant function can be thought of as a mapping GL(n,F)→Fx (=F\{0}) , this is then a homomorphism by the product rule for determinants.
When n>1, GL(n,F) is not an abelian group.
The special linear group SL(n,F) is the set of nXn matrices with determinant 1, or {g|g∈GL(n,F), det(g)=1}. It is an example of a normal subgroup of GL(n,F), since ∀h∈SL(n,F), ∀g∈GL(n,F) we have det(ghg-1)= det(g)det(h)det(g-1) = det(g)/det(g) = 1 (by various properties of determinant and SL(n,F) definition).