When you were small you learned that every whole number breaks into primes in exactly one way: 60 = 2^2 \cdot 3 \cdot 5, and no clever regrouping will ever give you a different set of prime factors. That fact — the Fundamental Theorem of Arithmetic — is so familiar that it feels like it must be true everywhere. It is the bedrock under fractions, greatest common divisors, and half of number theory.

Now that we have rings, the natural question is: does that theorem survive the leap to abstraction? Take some other integral domain — the Gaussian integers \mathbb{Z}[i], the polynomials \mathbb{Q}[x], the exotic \mathbb{Z}[\sqrt{-5}]. Do the "primes" of that ring still assemble numbers in one and only one way? The astonishing answer is: sometimes yes, sometimes no. The rings where unique factorization survives get a name of their own — unique factorization domains, or UFDs — and this page is about exactly where the line falls, and the famous ring where it snaps.

Two words that were the same in \mathbb{Z}

In \mathbb{Z} the word "prime" quietly does two jobs at once, and to generalise correctly we have to prise them apart. First, though, one piece of vocabulary we cannot do without: a unit is an element with a multiplicative inverse inside the ring. In \mathbb{Z} the only units are \pm 1; in a field every nonzero element is a unit. Units are the "trivial" factors — multiplying by one never counts as a real factorization, just as 7 = 1 \cdot 7 tells you nothing.

In \mathbb{Z} these two notions of "prime" coincide, which is why school never distinguishes them. In a general domain the relationship is one-directional: prime always implies irreducible, but the converse can fail. A UFD is precisely a ring civilised enough that the two words mean the same thing again.

The definition

The two clauses matter separately. Existence says factorizations don't run away forever (you always hit irreducibles and stop). Uniqueness says the destination doesn't depend on the path. Drop uniqueness and arithmetic loses its footing — greatest common divisors stop being well defined, and "the prime factorization" becomes "a prime factorization."

The ladder of nice rings

UFDs sit inside a tidy hierarchy of increasingly special integral domains. Each arrow is a genuine theorem, and — importantly — every arrow is strict: there are rings that clear one bar but not the next.

So the classics come for free: \mathbb{Z} (Euclidean, via the ordinary division algorithm), any polynomial ring F[x] over a field (Euclidean, via polynomial long division), and the Gaussian integers \mathbb{Z}[i] (Euclidean, via the norm) are all UFDs. But watch the strictness. There is a second, sideways source of UFDs that reaches rings the ladder misses:

Apply it with R = \mathbb{Z}: since \mathbb{Z} is a UFD, so is \mathbb{Z}[x]. Yet \mathbb{Z}[x] is not a PID — the ideal (2, x) needs two generators and can't be collapsed to one. So \mathbb{Z}[x] is the poster child for the missing arrow: UFD but not PID. Unique factorization is strictly weaker than "every ideal is principal."

Warm-up: factoring in the Gaussian integers

Before the famous disaster, a happy example. In \mathbb{Z}[i] the ordinary prime 5 is not irreducible — it splits:

How would you ever guess that? The trick that runs through this whole subject is the norm N(a + bi) = a^2 + b^2, which is multiplicative: N(zw) = N(z)N(w). It turns questions about factoring elements into questions about factoring ordinary integers. Here N(2 + i) = 4 + 1 = 5, a rational prime, so 2 + i can't be broken further — a nontrivial factor would have norm strictly between 1 and 5 dividing 5, and there is none. Both 2 \pm i are irreducible, and because \mathbb{Z}[i] is a UFD this is the factorization of 5, up to units (\pm 1, \pm i) and order.

The famous failure: \mathbb{Z}[\sqrt{-5}]

Now the ring that broke the dream. Work inside \mathbb{Z}[\sqrt{-5}] = \{\, a + b\sqrt{-5} : a, b \in \mathbb{Z} \,\}. It is a perfectly good integral domain — a subring of \mathbb{C}, no zero divisors. And yet the number 6 factors into irreducibles in two essentially different ways:

To be sure this is a real contradiction and not a trick of rearranging, we check two things with the norm N(a + b\sqrt{-5}) = a^2 + 5b^2 (again multiplicative). First, that all four factors are irreducible. Their norms are

Crucially, no element of \mathbb{Z}[\sqrt{-5}] has norm 2 or 3: the equation a^2 + 5b^2 = 2 or = 3 has no integer solutions (if b \neq 0 the norm is at least 5, and a^2 = 2, 3 fails). So if, say, 2 = \alpha\beta then N(\alpha)N(\beta) = 4 would force one factor to have norm 1 — a unit. Hence 2 is irreducible, and the same norm argument handles 3 and 1 \pm \sqrt{-5}.

Second, that the two factorizations are not just associates in disguise. The units of \mathbb{Z}[\sqrt{-5}] are only \pm 1 (a unit has norm 1, and a^2 + 5b^2 = 1 forces b = 0, a = \pm 1). Since 2 \neq \pm(1 \pm \sqrt{-5}), the factor 2 is an associate of none of the right-hand factors. The two factorizations are irreducibly, genuinely distinct. Unique factorization is dead in this ring.

Where prime and irreducible part ways

The failure above is the same coin, other side: in \mathbb{Z}[\sqrt{-5}] there is an irreducible element that is not prime. Look at 2. We just proved it is irreducible. But is it prime? It divides the product

so 2 \mid (1 + \sqrt{-5})(1 - \sqrt{-5}). If 2 were prime it would have to divide one of the two factors. But \frac{1 + \sqrt{-5}}{2} = \tfrac12 + \tfrac12\sqrt{-5} is not in the ring (its coordinates aren't integers), and likewise for 1 - \sqrt{-5}. So 2 \nmid (1 + \sqrt{-5}) and 2 \nmid (1 - \sqrt{-5}): 2 is irreducible but not prime. That is exactly why factorization can fork — irreducibles that lack the divides-a-product property don't have to line up between rival factorizations.