algebra

Modules

Here’s a truth that tidied up half of algebra for me: a module is a vector space where the scalars are only allowed to come from a ring, so you can scale but not always unscale.

Change that one word, field to ring, and a surprising amount of new behaviour appears.

The core idea, in plain words

Take a Ring $R$ (think of the integers $\mathbb{Z}$ ). An $R$ -module is an abelian Group $(M, +)$ together with a way to multiply elements of $M$ by elements of $R$ , written $r \cdot m$ , such that for all $r, s \in R$ and $m, n \in M$ :

$r(m + n) = rm + rn$
$(r + s)m = rm + sm$
$(rs)m = r(sm)$
$1 \cdot m = m$

If $R$ is a Field, this is word for word the definition of a Vector Space. So modules generalise vector spaces. They also generalise abelian groups: every abelian group is automatically a $\mathbb{Z}$ -module, with $3 \cdot m$ meaning $m + m + m$ and $(-2) \cdot m$ meaning $-(m + m)$ . And they generalise ideals: an ideal of $R$ is just a module sitting inside $R$ itself.

That’s the beauty of it. Three ideas from three pages turn out to be one idea.

An everyday analogy: buying in packs

Imagine a shop that only sells things in whole packs. You can buy 3 packs or 7 packs, you can return packs (negative amounts), but you can never buy half a pack. Your shopping basket is a module over $\mathbb{Z}$ : you can scale it by any whole number, but “divide the basket by 2” might have no answer. A vector space is a deli counter where you can ask for any amount by weight. A module is the packaged-goods aisle.

A worked example: the 6-hour clock as a $\mathbb{Z}$ -module

Take $M = \mathbb{Z}_6 = \{0, 1, 2, 3, 4, 5\}$ under clock addition. As a $\mathbb{Z}$ -module, $n \cdot m$ means “add $m$ to itself $n$ times, mod 6”.

$4 \cdot 5 = 20 \equiv 2$ .
$6 \cdot m = 0$ for every $m$ . Something non-zero in the ring kills the whole module.

That never happens in a vector space: if $c \neq 0$ is a scalar and $cv = 0$ , multiply by $c^{-1}$ to get $v = 0$ . But in a module you can’t divide, so non-zero scalars can kill non-zero elements. Such elements are called torsion.

Now try to find a basis of $\mathbb{Z}_6$ , a set of elements that every element is a unique combination of with integer coefficients. Try $\{1\}$ : every element is $n \cdot 1$ , but not uniquely, since $0 \cdot 1 = 6 \cdot 1$ . In fact $\mathbb{Z}_6$ has no basis at all. Modules that do have a basis are called free, like $\mathbb{Z}^2$ with basis $(1,0), (0,1)$ . Every vector space is free. Most modules are not.

The crowning result for nice rings like $\mathbb{Z}$ is the structure theorem: every finitely generated abelian group looks like

\mathbb{Z}^r \oplus \mathbb{Z}_{n_1} \oplus \mathbb{Z}_{n_2} \oplus \dots \oplus \mathbb{Z}_{n_k}

a free part plus a torsion part. For instance $\mathbb{Z}_6 \cong \mathbb{Z}_2 \oplus \mathbb{Z}_3$ .

A picture in words: a lattice in the plane

Picture the plane $\mathbb{R}^2$ , a vector space: you can stretch any arrow by any real amount. Now keep only the points with integer coordinates, a square grid of dots like an orchard. That’s $\mathbb{Z}^2$ , a free $\mathbb{Z}$ -module. You can add dots and scale by whole numbers, but halving the dot $(1, 1)$ lands you between the trees.

Now stand at the origin and choose a sub-orchard: only the dots $(a, b)$ with $a$ and $b$ both even. Squash the orchard by it, as in Quotient Groups, and what’s left is four classes of dots: $\mathbb{Z}_2 \oplus \mathbb{Z}_2$ , a little torsion module.

graph TD
  M["Modules over a ring R"] --> V["R a field: vector spaces"]
  M --> A["R = ℤ: abelian groups"]
  M --> I["Submodules of R: ideals"]

Common misconceptions

“Every module has a basis.” Only free ones. $\mathbb{Z}_6$ and $\mathbb{Q}$ (as $\mathbb{Z}$ -modules) have none.
“Bases all have the same size, so dimension is always defined.” For free modules over commutative rings, yes. Over strange non-commutative rings, even that can fail.
“A subset that spans and is minimal is a basis.” In vector spaces, yes. In $\mathbb{Z}$ , the set $\{2, 3\}$ spans (since $3 - 2 = 1$ ) and removing either element breaks it, yet it is not a basis, because $3 \cdot 2 - 2 \cdot 3 = 0$ is a non-trivial relation.
“Modules are an exotic abstraction.” They are everywhere: a Linear Operator $T$ on a vector space turns it into a module over the polynomial ring $k[x]$ , with $x$ acting as $T$ . Jordan normal form is the structure theorem applied to that module.

How it connects

Down to Rings, Commutative Rings and Groups: a module is a ring acting on an abelian group, just as in Group Actions a group acts on a set.
Sideways to Fields: over a field, every module is a free vector space.
Up to Vector Spaces, Linear Operators and Matrices and Tensors, where tensor products are defined for modules in general.
Out to Category Theory, whose early life was spent studying categories of modules.

The truth revealed: a vector space is a module that got lucky. Take away division, and you find out how much of linear algebra was luck.