Let free Abelian groupG and G′ have bases a1,⋯,an and a1′,⋯,am′, respectively. If function f:G→G′ is a homomorphism, then there exists a unique set of integers {λij}⊂Z satisfying the following.
f(aj)=i=1∑mλijai′
The matrix (λij)∈Zm×n is referred to as the matrix of f (with respect to the bases of G and G′).
If free Abelian groups G and G′ have ranks n,m, and f:G→G′ is a homomorphism, then there exists a homomorphism g with the following matrix.
d1000⋮00⋱00⋮000dr0⋮00000⋮0⋯⋯⋯⋯⋱⋯0000⋮0∈Zm×n
Here, d1,⋯,dr∈N and d1∣⋯∣dr, meaning dk has to be a divisor of dk+1.
Description
The matrix mentioned in the theorem is known as the Smith normal form, given f and λij, the d1,⋯,dr can be found using the Gauss elimination method and thus are a linear combination of λij. Row operations involved in obtaining the Smith normal form affect the basis of G′, and column operations affect the bases of G.
In essence, this theorem suggests that when considering two free groupsG,G′, it is sufficient to have only r≤min(m,n) of the d1,⋯,dr instead of all λij. These can be seen as the most succinct summary of information from G to G′.
f(a)=f(b)=x+y−zx−y+z
For example, if f:F[a,b]→F[x,y,z] is defined as above, it is naturally a homomorphism and the matrix of f is as follows.
[111−1−11]∼[100200]
The right side is the Smith normal form of the left side. Once a homomorphism’s matrix takes on the form of the Smith normal form, that form is unique.
Theorem
Choose any bases for G and G′, and define any homomorphism f(aj)=∑i=1mλijai′. The matrix of f, (λij), belongs to the set of matrices made of integers Zm×n.
Since Z is a principal ideal domain, for matrix (λij), there exists a Smith normal form where all elements except for r diagonal elements d1,⋯,dr are 0.
■
Munkres. (1984). Elements of Algebraic Topology: p55. ↩︎