Modules over a PID

Overview

Modules over a principal ideal domain (PID) — rings like ℤ or K[x] where every ideal is generated by one element — enjoy a structure theory almost as clean as vector spaces. The Structure Theorem for Finitely Generated Modules over a PID says every such module splits as a direct sum of a free part and finitely many cyclic torsion pieces, each a quotient R/(p^k) for a prime p. Specializing R = ℤ recovers the classification of finite abelian groups; specializing R = K[x] recovers Jordan/rational canonical form for a linear operator. This single theorem is therefore the common ancestor of two results usually taught separately.

Intuition

Over a field, every finitely generated module (= vector space) is free — it's just Rⁿ, no exceptions. Over a general PID, a finitely generated module can have 'twisted' pieces that aren't free, but the twisting is completely tamed: every such piece looks like R/(pᵏ) for a single prime p and exponent k, exactly the building block you'd get from long division with remainder. The free rank n counts the 'genuinely infinite' directions, while the torsion summands R/(pᵏ) record finite, prime-power-sized obstructions — for R=ℤ these are literally the cyclic groups ℤ/pᵏℤ appearing in a finite abelian group's decomposition, and for R=K[x] they are exactly the Jordan blocks (for p = x−λ) that classify a linear map up to conjugation.

Formal Definition

Definition

Let R be a PID and M a finitely generated R-module. Then M decomposes (uniquely up to the listed data) as:

M \cong R^n \oplus R/(p_1^{e_1}) \oplus R/(p_2^{e_2}) \oplus \cdots \oplus R/(p_k^{e_k})

Primary decomposition (invariant factor / elementary divisor form)

n = \operatorname{rank}(M), \quad p_i \text{ prime in } R, \quad e_i \geq 1

n is the free rank; the p_i^{e_i} are the elementary divisors

M \cong R^n \oplus R/(d_1) \oplus \cdots \oplus R/(d_s), \quad d_1 \mid d_2 \mid \cdots \mid d_s

Equivalent invariant factor form: d_1 | d_2 | ... | d_s

\operatorname{Tor}(M) = \{ m \in M : rm = 0 \text{ for some nonzero } r \in R \}

Torsion submodule

Properties

Free submodules

\text{A submodule of a free module over a PID is again free (of rank} \le \text{the ambient rank).}

Torsion-free = free (f.g. case)

\text{A finitely generated torsion-free module over a PID is free.}

Uniqueness of invariants

\text{The rank } n \text{ and the multiset of elementary divisors } p_i^{e_i} \text{ are uniquely determined by } M.

Classification of finite abelian groups

R=\mathbb{Z}: \ M \cong \mathbb{Z}/(p_1^{e_1}) \oplus \cdots \oplus \mathbb{Z}/(p_k^{e_k}) \text{ for a finite abelian group } M \text{ (rank } n=0\text{).}

Rational/Jordan canonical form

R=K[x]: \ \text{a } K[x]\text{-module structure on } V \text{ (via a linear operator } T\text{) decomposes into } K[x]/(p(x)^e) \text{ pieces, giving canonical forms for } T.

Worked Examples

A finite abelian group is a torsion ℤ-module (rank n=0); its order factors as 12 = 2²·3, so its elementary divisors are prime powers dividing 2² and 3.
$|M| = 12 = 2^2 \cdot 3$
For the prime 2 with exponent sum 2, the partitions of 2 give either one factor ℤ/4 or two factors ℤ/2 ⊕ ℤ/2.
$\mathbb{Z}/4 \quad \text{or} \quad \mathbb{Z}/2 \oplus \mathbb{Z}/2$
For the prime 3 with exponent sum 1, there is only one option, ℤ/3. Combine each 2-part choice with the 3-part.
$M \cong (\mathbb{Z}/4 \oplus \mathbb{Z}/3) \ \text{or}\ (\mathbb{Z}/2 \oplus \mathbb{Z}/2 \oplus \mathbb{Z}/3)$

Answer: There are exactly two abelian groups of order 12 up to isomorphism: ℤ/4 ⊕ ℤ/3 (≅ ℤ/12) and ℤ/2 ⊕ ℤ/2 ⊕ ℤ/3 (≅ ℤ/2 ⊕ ℤ/6).

Practice Problems

Difficulty 6/10

List all abelian groups of order 8 up to isomorphism.

Difficulty 7/10

Why is every finitely generated torsion-free module over a PID automatically free? Give the intuition, not a full proof.

Difficulty 8/10

Explain how Jordan canonical form for a linear operator T on a finite-dimensional vector space V over ℂ arises as a special case of the structure theorem, taking R = K[x].

Common Mistakes

Common Mistake

Assuming the structure theorem applies over any ring, not just a PID.

The theorem crucially uses that R is a PID (so every submodule of a free module is free); over a general ring, modules can fail to decompose this cleanly.