In the integers you take one fact so utterly for granted that it feels like a law of nature: if a product is zero, one of the factors must be zero. ab = 0 forces a = 0 or b = 0. It is the reason you are allowed to solve (x - 2)(x + 3) = 0 by setting each bracket to zero — the whole machinery of factoring rests on it.

But a ring need not obey it. On a 6-hour clock, 2 \times 3 = 6 \equiv 0: two perfectly nonzero numbers multiply to nothing. That single crack — a product of nonzeros collapsing to zero — breaks cancellation, wrecks factoring, and separates the well-behaved rings from the wild ones. The rings that forbid the crack are the integral domains; the ones where you can also always divide are the fields. This page is about those two families and the beautiful theorem that ties them to the prime numbers.

Zero divisors: the crack in the ring

A zero divisor is a nonzero element that can multiply another nonzero element to give zero. Formally, in a commutative ring R, an element a \neq 0 is a zero divisor if there exists some b \neq 0 with ab = 0.

In modular arithmetic mod 6, the culprits are exactly 2, 3, 4:

Each of these is a nonzero number whose "shadow" reaches zero. Compare \mathbb{Z}, where this can never happen: the product of two nonzero integers is always nonzero. That single difference is what we are about to promote into a definition — and the table below shows you exactly where \mathbb{Z}_6 goes wrong.

Read the highlighted cells off the grid: they sit exactly at rows and columns 2, 3, 4 — the numbers that share a common factor with 6. The numbers 1 and 5 never appear in a bad cell, because they are coprime to 6. Hold that thought: it is the seed of the whole classification.

The definition of an integral domain

The prototype, of course, is \mathbb{Z} itself — the word "integral" is a nod to the integers. Also domains: \mathbb{Q}, \mathbb{R}, \mathbb{C}, every polynomial ring \mathbb{R}[x], and \mathbb{Z}_p for prime p. Not domains: \mathbb{Z}_6 (we just watched it fail) and the 2 \times 2 matrices (non-commutative, and stuffed with zero divisors like \begin{bmatrix}1&0\\0&0\end{bmatrix}\begin{bmatrix}0&0\\0&1\end{bmatrix} = 0).

The payoff of banning zero divisors is the cancellation law, the thing that makes a domain feel like ordinary arithmetic:

The proof is one line and shows exactly where the no-zero-divisors axiom earns its keep. From ab = ac we get a(b - c) = 0. Since a \neq 0 and there are no zero divisors, the other factor must vanish: b - c = 0, i.e. b = c. Notice we never divided by a — cancellation is weaker than division, and it is all a domain promises. In \mathbb{Z}_6 cancellation already fails: 2 \cdot 1 = 2 \cdot 4 = 2, yet 1 \neq 4.

Units, and the definition of a field

A domain forbids the bad behaviour of zero divisors. A field goes further and demands the good behaviour of full division. The gatekeeper is the idea of a unit.

A unit is an element u that has a multiplicative inverse — some u^{-1} in the ring with u\,u^{-1} = 1. In \mathbb{Z} the only units are \pm 1 (nothing else has an integer reciprocal). In \mathbb{Z}_8 the units are 1, 3, 5, 7 — exactly the residues coprime to 8 — since 3 \cdot 3 = 9 \equiv 1 and 5 \cdot 5 = 25 \equiv 1 and 7 \cdot 7 = 49 \equiv 1 \pmod 8.

The familiar fields are \mathbb{Q}, \mathbb{R}, \mathbb{C} — every nonzero rational, real or complex number has a reciprocal. The integers \mathbb{Z} are not a field, because 2 has no inverse inside \mathbb{Z}. And the finite fields \mathbb{Z}_p (with p prime) are the stars of the next section.

Worked example: ℤ₅ is a field

Take \mathbb{Z}_5 = \{0, 1, 2, 3, 4\}. To prove it is a field we only need to find an inverse for each nonzero element — a partner that multiplies with it to 1 \pmod 5:

Every nonzero residue is paired off, so every nonzero element is a unit — and \mathbb{Z}_5 is a field. Contrast \mathbb{Z}_6: try to invert 2 and you run through 2\cdot1, 2\cdot2, \dots, 2\cdot5 = 2, 4, 0, 2, 4 and never hit 1. The value 2 is not a unit — it is a zero divisor instead, and no zero divisor can ever be a unit (if ab = 0 with b\neq0 and a had an inverse, then b = a^{-1}ab = a^{-1}\cdot 0 = 0, a contradiction).

This is the heart of the matter. In \mathbb{Z}_n, an element a is a unit exactly when \gcd(a, n) = 1, and it is a zero divisor exactly when \gcd(a, n) > 1 (and a\neq0). Every nonzero residue is one or the other. So all nonzero elements are units precisely when every number below n is coprime to it — which happens if and only if n is prime.

The theorem that ties it together

If n = ab is composite with 1 < a, b < n, then a and b are nonzero in \mathbb{Z}_n but ab = n \equiv 0 — instant zero divisors, so no domain. If n = p is prime, every nonzero residue is coprime to p, hence invertible, hence \mathbb{Z}_p is a field. And a field is automatically a domain (proven above: units can't be zero divisors). The three statements march in lockstep.

That a domain and a field coincide here is not an accident of small numbers. It is a genuine theorem about finite rings:

The idea: fix a nonzero a in a finite domain and look at the map x \mapsto ax. By the cancellation law it is injective (if ax = ay then x = y), and an injective map from a finite set to itself must be surjective. So something maps to 1: there is an x with ax = 1. That x is the inverse of a — every nonzero element is a unit, and the domain is a field. The finiteness is doing all the work; \mathbb{Z} is an infinite domain that escapes into no field.

The full hierarchy

Stacking the definitions gives a clean chain of strictly nested families. Reading from most permissive to most demanding:

Every field is an integral domain — division is more than enough to kill zero divisors. But the reverse fails, and \mathbb{Z} is the standard counterexample: a perfectly good domain with no zero divisors at all, yet 2 has no inverse, so it is not a field. The gap between "no zero divisors" and "always invertible" is exactly the gap between a domain and a field — and, as we just saw, that gap closes the moment the ring is finite.