What a Group Is, and Why the Axioms Are So Bare

Ask what makes a square symmetric and the useful answer is not "it looks the same on both sides". It is that there is a list of things you can do to it which leave it looking identical: rotate a quarter turn, flip it about a diagonal, and so on.

That shift, from symmetry as an adjective to symmetry as a collection of actions, is what makes it mathematics. And the collection has structure of its own. Do two symmetries in a row and you get a symmetry. Every symmetry can be undone. Doing nothing is a symmetry.

Key idea: Groups are what you get when you take that observation seriously. The subject studies the set of allowed actions rather than the object they act on, which is why the same theorem covers a crystal, a shuffle, and an error-correcting code.

A group is a set $G$ with one operation, written as juxtaposition, satisfying exactly four conditions:

1. Closure. $a, b \in G \implies ab \in G$. Combining two elements never leaves the set.
2. Associativity. $(ab)c = a(bc)$. Grouping does not matter.
3. Identity. There is an $e$ with $ea = ae = a$ for all $a$. Doing nothing is available.
4. Inverses. Every $a$ has an $a^{-1}$ with $a a^{-1} = e$. Everything is undoable.

Notice what is absent. No requirement that $ab = ba$. No second operation. No numbers, no ordering, no notion of size or distance.

Definition: A group is a set with an associative operation that has an identity and in which every element is invertible. A group where $ab = ba$ always holds is called abelian, after Niels Abel, who appears again at the end of this course.

Number the corners 1 to 4. Every symmetry of the square permutes them, and there are exactly eight:

This is the dihedral group $D_4$, of order 8. Every element is built from two generators, one rotation and one reflection, and the whole group collapses into two relations: $r^4 = e$ and $s^2 = e$, plus one more that says how they interact.

The order of an element is how many times you apply it to get back to the identity. Rotating by 90 degrees four times returns the square to start, so $r$ has order 4.

That failure of commutativity is not an awkwardness to be worked around. It is the source of nearly everything interesting in group theory. Abelian groups are, broadly, well understood and finitely classified. Non-abelian groups contain the whole subject: the symmetric groups, the Rubik's cube, the symmetries of physical law, and the obstruction that kills the quintic formula.

The XOR row is worth pausing on. Every bit string is its own inverse, since $a \oplus a = 0$, which is exactly why XOR encryption decrypts with the same operation and why XOR-based checksums cancel cleanly.

The integers under multiplication are not a group: 2 has no integer inverse. One missing axiom is enough to disqualify.

A single element generates everything you can reach by repeating it. In $\mathbb{Z}_{12}$ under addition, starting from 5 gives $5, 10, 3, 8, 1, 6, 11, 4, 9, 2, 7, 0$: all twelve. Starting from 4 gives only $4, 8, 0$.

def generated(start, n):
    seen, x = [], 0
    while True:
        x = (x + start) % n
        seen.append(x)
        if x == 0:
            return seen

generated(5, 12)   # 12 elements: 5 generates all of Z_12
generated(4, 12)   # 3 elements:  4, 8, 0

A group generated by one element is cyclic. Which elements generate $\mathbb{Z}_n$? Exactly those coprime to $n$, which is the same condition that governed modular inverses in the discrete mathematics course. That is not a coincidence, and the next lesson explains why the two facts are the same fact.

A regular $n$-gon has $2n$ symmetries: $n$ rotations and $n$ reflections. In three dimensions the counts jump, because rotation axes multiply.

The cube and the octahedron both have 48, and the dodecahedron and icosahedron both have 120. That is not an accident of counting: each pair is dual, one's vertices sitting at the other's face centres, so a symmetry of either is a symmetry of both. Two objects that look nothing alike have literally the same symmetry group, which is exactly the kind of statement the bare definition was built to support.

Take the moves of a Rubik's cube as elements: closure holds (a sequence of moves is a move), associativity holds, doing nothing is the identity, and every sequence can be undone. It is a group, and its order can be counted directly:

$|G| = \frac{8! \cdot 3^7 \cdot 12! \cdot 2^{11}}{2} = 43{,}252{,}003{,}274{,}489{,}856{,}000$

The $8!$ and $12!$ are arrangements of corners and edges; the powers of 3 and 2 are their orientations, each with one degree of freedom removed because the last piece's twist is forced; the final division by 2 is because corner and edge permutations cannot have opposite parity.

In practice: Every one of those constraints is a group-theoretic fact, and each is why certain "impossible" cube states exist. A single twisted corner or a single swapped pair of edges cannot be reached by legal moves, which is how you can tell a cube has been prised apart and reassembled rather than scrambled.

A stronger definition would prove more about fewer things. The four axioms were chosen to be the smallest set that still supports substantial theorems, and the payoff is reach.

When you prove that every element of a finite group has an order dividing the group's size, you have simultaneously proved a fact about clock arithmetic, a fact about card shuffles, a fact about crystal symmetries, and Fermat's little theorem. Nobody has to notice the connection; it is forced by the axioms.

Key idea: Abstraction here is not decoration. It is a compression scheme for theorems. The cost is that a group on its own tells you almost nothing about the object it came from, so the practical skill is recognising when a situation is a group, and then reading the results off the shelf.

The next lesson proves the theorem just described, and the one after it shows how physics and chemistry read their results off that same shelf.

Given a set and an operation, four questions settle it, and the failures are more informative than the successes.

1. Is it closed? Odd integers under addition fail immediately: two odds sum to an even.
2. Is it associative? Subtraction fails: $(5 - 3) - 1 = 1$ but $5 - (3 - 1) = 3$. Function composition and matrix multiplication always pass.
3. Is there an identity? Positive integers under addition have none, since 0 is excluded.
4. Does every element have an inverse? This is the usual failure. Integers under multiplication, square matrices under multiplication, and strings under concatenation all fall here.

Structures that fail only the last test are common enough to have their own name, monoids, and they are what you have whenever an operation composes and has a unit but cannot be undone: string concatenation, list append, and the merge step of a MapReduce.