Advanced Algebra II
modules over principal ideal domain
In order to prove the uniqueness of the decomposition and also an- other description by invariant factors. We need to work more.
Lemma 0.1. Let M, N be module over a principal ideal domain, r ∈ R and p ∈ R is prime.
(1) rM and M[r] := {x ∈ M|rx = 0} are submodules.
(2) M[p] is a vector space over R/(p).
(3) (R/(pn))[p] ∼= R/(p) and pm(R/(pn)) ∼= R/(pn−m) for m < n.
(4) If M ∼= ⊕Mi, then rM ∼= ⊕rMi and M[r] ∼= ⊕Mi[r].
(5) If f : M → N a RM-isomorphism, then f : Mτ ∼= Nτ, and f : M(p) ∼= N(p).
Proof. The proof are straightforward. We leave it to the readers. ¤ Lemma 0.2. Let r =Q
i=1,...,kpnii. Then
R/(r) ∼= ⊕i=1,...,kR/(pnii).
Proof. The is a generalization of Chinese Remainder Theorem. Just
copy the proof then we are done. ¤
Theorem 0.3. Let M be a finitely generated module over a principal ideal domain R. Then
(1) M is a direct sum of a free module F of finite rank and a finite number of cyclic torsion modules of order r1, ..., rm respectively.
Where r1|r2|...|rm. And the rank of F and the list of ideals (r1), ..., (rm) is unique.
(2) M is a direct sum of a free module F of finite rank and a fi- nite number of cyclic torsion modules of order pa11, ..., pakk respec- tively. And the rank of F and the list of ideals (pa11), ..., (pakk) is unique.
The r1, ..., rm are called invariant factors. And pa11, ..., pakk are called elementary divisors.
Proof. We have seen the existence of decomposition into elementary divisors. By the similar method as we did in finitely generated abelian groups, one can construct the invariant factors out of elementary divi- sors. Thus it suffices to prove uniqueness for both cases.
For a fix prime p, we consider M[p]. It’s a vector space over R/(p) and d1 := dimR/(p)M[p] measure the number of elementary divisor of order pn. We next consider pM[p] and define d2 := dimR/(p)pM [p].
Then d2 measure the number of elementary divisor of order pn with n ≥ 2. Inductively, we define dk := dimR/(p)pk−1M[p] which measure the number of elementary divisor of order pn with n ≥ k. The point is that these dk are uniquely determined by M. It’s easy to see that M
1
2
has dk− dk+1 elementary divisor (pk), and this is uniquely determined
by M. ¤
Corollary 0.4 (Jordan canonical form). Let A be a n × n matrix over a field k. Suppose that it satisfies a polynomial (in particular, char- acteristic polynomial or minimal polynomial) which splits into linear factors. Then A has Jordan canonical form.
Proof. We view A as a linear transformation on a n-dimensional vector space V over k. Then V can be viewed as a k[x]-module by f (x)v :=
f (A)v. It’s well-known that k[x] is a principal ideal domain. It’s clear that V is a torsion module since A satisfies a polynomial. Thus V = PV (x − λ), or even V decomposes into cyclic submodule of order (x − λi)ki. It’s easy to check that submodule of V is a vector subspace.
And one can pick a basis in a cyclic submodule so that the matrix is
represented as a Jordan block. ¤
Corollary 0.5 (fundamental theorem of finitely generated abelian group).