Polynomial Factorization over UFDs

Suppose you know a ring RR has unique factorization. Every nonzero nonunit splits into irreducibles in an essentially unique way. Does this property survive when you build the polynomial ring R[x]R[x]? This question is not idle curiosity. It is the structural fact underlying why Z[x]\mathbb{Z}[x], k[x,y]k[x,y], and countless coordinate rings of algebraic varieties all admit unique factorization. This holds even though none of these rings is a Principal Ideal Domain or Euclidean domain. The naive hope is that factoring a polynomial over RR should be no harder than factoring it over the fraction field KK of RR. There, the machinery of a field (division, degree arguments, standard polynomial-ring theorems) is fully available. The subtlety is that a polynomial can factor in K[x]K[x] using denominators. These denominators do not actually obstruct an integral factorization in R[x]R[x]. This is exactly the content of Gauss's Lemma, which lets us clear denominators without introducing new irreducible factors. The theorem in this article, due in essence to Gauss, promotes that lemma into a complete structural statement. If RR is a Unique Factorization Domain, then so is R[x]R[x]. Moreover, so is R[x1,…,xn]R[x_1,\dots,x_n] for every nn. The big picture is a bridge between two worlds of factorization: the "constant" factorizations happening inside RR, and the "polynomial" factorizations happening inside K[x]K[x]. We show these combine seamlessly, with no interference, to produce unique factorization in R[x]R[x]. This is one of the most important closure properties in commutative algebra. It is what makes multivariable polynomial rings over Z\mathbb{Z} or over a field tractable. It also lays the groundwork for irreducibility tests such as Eisenstein's Criterion, which detect irreducibility in R[x]R[x] by reducing mod a prime of RR. This article gives the precise definitions, and proves the theorem completely (via primitive polynomials and content). It works two nontrivial examples, and shows exactly which hypothesis is essential. It does this by exhibiting a UFD-like failure when that hypothesis is dropped.

Key takeaways
  • A UFD is an integral domain where every nonzero nonunit factors into irreducibles, uniquely up to order and units.
  • Gauss's Lemma says a product of primitive polynomials in R[x]R[x] is primitive, and controls how factorizations move between R[x]R[x] and K[x]K[x].
  • The content c(f)c(f) of a polynomial is the gcd of its coefficients; every f∈R[x]f \in R[x] is c(f)c(f) times a primitive polynomial.
  • Main theorem: if RR is a UFD then R[x]R[x] is a UFD, and by induction R[x1,…,xn]R[x_1,\dots,x_n] is a UFD for all n≥1n \ge 1.
  • Irreducibles in R[x]R[x] are exactly: irreducibles of RR, together with primitive polynomials of positive degree that are irreducible in K[x]K[x].
  • Key idea: unique factorization in R[x]R[x] decomposes cleanly into unique factorization of contents in RR and unique factorization of primitive parts in K[x]K[x].

Where this lives today

The fact that k[x1,…,xn]k[x_1,\dots,x_n] is a UFD underlies the theory of algebraic varieties: irreducible polynomials cut out irreducible hypersurfaces, and unique factorization lets geometers speak unambiguously of the "degree" and "components" of a variety defined by a single equation. In algebraic number theory, Z[x]\mathbb{Z}[x] being a UFD is the starting point for factoring minimal polynomials and analyzing ramification via reduction mod primes — the same reduction trick used in the proof of Gauss's Lemma reappears as the key tool behind Eisenstein's Criterion, a fast test for irreducibility of integer polynomials that is indispensable for proving polynomials like cyclotomic polynomials are irreducible over Q\mathbb{Q}. In cryptography, polynomial factorization over finite fields (themselves UFDs, indeed fields) and over Z\mathbb{Z} underlies the design and cryptanalysis of lattice-based and code-based schemes, where distinguishing reducible from irreducible polynomials mod small primes is a practical computational step. In computer algebra systems, factorization algorithms for Z[x]\mathbb{Z}[x] (such as those based on Hensel lifting) exploit exactly the content/primitive-part split proven here, first factoring modulo a prime and then lifting, using Gauss's Lemma to certify that the lifted factorization is genuine over Z\mathbb{Z}.

:::quiz[Which of the following best describes the irreducibles of R[x] when R is a UFD?]

  • Only the irreducibles of R
  • Only the primitive polynomials irreducible in K[x]
  • Irreducibles of R together with primitive positive-degree polynomials irreducible in K[x] ✓
  • All polynomials irreducible in K[x], regardless of primitivity

:::quiz[In Z[x], the polynomial 4x + 6 is]

  • irreducible in Z[x]
  • primitive
  • reducible in Z[x], since 4x+6 = 2(2x+3) with both factors nonunits ✓
  • irreducible in Q[x] and hence automatically irreducible in Z[x]

:::quiz[Why does Z[√-5][x] fail to be a UFD?]

  • Because Z[√-5] is not an integral domain
  • Because Z[√-5] itself is not a UFD, and this failure persists in degree 0 of the polynomial ring ✓
  • Because Z[√-5][x] has no irreducible elements
  • Because Q(√-5)[x] is not a PID

:::quiz[The proof that R[x] is a UFD when R is a UFD relies critically on which prior fact about K[x], K = Frac(R)?]

  • That K[x] has no irreducible elements
  • That K[x] is a UFD (in fact a PID via the division algorithm) ✓
  • That K[x] equals R[x]
  • That K[x] has finitely many units

:::faq[Is R[x] a PID whenever R is a UFD?] No. The theorem only promotes unique factorization, not the stronger property of every ideal being principal. The standard counterexample is R=ZR = \mathbb{Z}: Z[x]\mathbb{Z}[x] is a UFD (by this theorem) but not a PID, since the ideal (2,x)(2,x) consisting of polynomials with even constant term is not principal — no single polynomial generates both 22 and xx. More generally, R[x]R[x] is a PID if and only if RR is a field, because if RR is not a field, take any nonzero nonunit a∈Ra \in R; the ideal (a,x)(a,x) in R[x]R[x] fails to be principal by a degree/content argument similar to the Z[x]\mathbb{Z}[x] case.

:::faq[What exactly is the "content" when R is not Z, e.g. R = k[y]?] The content is defined identically: a greatest common divisor, computed inside RR, of all the coefficients of the polynomial, viewed as a polynomial in the remaining variable(s). For R=k[y]R=k[y] and f(x,y)f(x,y) a polynomial in xx with coefficients in k[y]k[y], the content is the gcd (in k[y]k[y], using that k[y]k[y] is itself a UFD/PID) of those coefficient polynomials in yy. This is exactly what Example 2 computes, extracting c(f)=y(y−1)c(f)=y(y-1) as a common factor of all coefficients.

:::faq[Does the theorem extend to formal power series Rx?] Yes, with a different proof: if RR is a UFD, then R[[x]]R[[x]] is also a UFD, but this requires separate arguments (using that units of R[[x]]R[[x]] are power series with unit constant term, and a different notion of "order" replacing degree) and is not a direct corollary of the polynomial-ring theorem proved here. It is a related but logically independent result.

:::faq[Why can't we just say "K[x] is a UFD, so done" and skip R entirely?] Because factoring in K[x]K[x] only tells you about factorization up to units of KK, which is a much coarser equivalence than factorization up to units of RR. Two elements of R[x]R[x] can be associates in K[x]K[x] (differ by a nonzero scalar in KK) without being associates in R[x]R[x] (differing only by a unit of RR) — for instance 2x2x and xx are associates in Q[x]\mathbb{Q}[x] but not in Z[x]\mathbb{Z}[x]. The content machinery is precisely what reconciles the two notions of "sameness."

Ask the tutor: How does Gauss's Lemma get used to prove Eisenstein's Criterion, and why does Eisenstein's test require checking primitivity first?

DefinitionUnique Factorization Domain, content, and primitive polynomial

Let RR be an integral domain.

UFD. RR is a unique factorization domain if every nonzero nonunit a∈Ra \in R can be written as a product of irreducible elements

a=p1p2⋯pn,a = p_1 p_2 \cdots p_n,

and this factorization is unique up to reordering and multiplication of the pip_i by units: if also a=q1⋯qma = q_1 \cdots q_m with each qjq_j irreducible, then n=mn = m and after reindexing qi=uipiq_i = u_i p_i for units ui∈R×u_i \in R^\times.

Content. Assume now RR is a UFD with fraction field KK. For a nonzero polynomial

f=a0+a1x+⋯+anxn∈R[x],f = a_0 + a_1 x + \cdots + a_n x^n \in R[x],

the content c(f)c(f) is a greatest common divisor of a0,…,ana_0,\dots,a_n in RR (well-defined up to a unit, since RR is a UFD and gcds exist and are unique up to units). We say ff is primitive if c(f)c(f) is a unit, i.e. no irreducible of RR divides all coefficients of ff. Every nonzero f∈R[x]f \in R[x] factors as f=c(f) f1f = c(f)\, f_1 with f1∈R[x]f_1 \in R[x] primitive; f1f_1 is called a primitive part of ff, well-defined up to a unit of RR. An element of R[x]R[x] is irreducible if it is a nonzero nonunit of R[x]R[x] that cannot be written as a product of two nonunits of R[x]R[x]. (Note: the units of R[x]R[x] are exactly the units of RR, since RR is a domain.)

Intuition

Think of a polynomial in R[x]R[x] as having two independent "layers" of structure: a scalar layer, given by the content living in RR, and a shape layer, given by the primitive part, whose factorization behavior mimics factorization in the honest field K[x]K[x]. Gauss's Lemma is the statement that these two layers never interact destructively: multiplying two primitive polynomials never manufactures a nontrivial common scalar factor in the product's coefficients. This is the miracle that makes the theorem work. In K[x]K[x] we have every tool of a Euclidean domain: division algorithm, degree induction, unique factorization automatically (since K[x]K[x] is even a PID). The danger in trying to transport this to R[x]R[x] is that an irreducible factorization over K[x]K[x] might use coefficients with denominators — for instance 2x2−82x^2 - 8 factors as 2(x−2)(x+2)2(x-2)(x+2) in Z[x]\mathbb{Z}[x], but you might worry a subtler polynomial could only be split using fractions that do not clear integrally. Gauss's Lemma guarantees this never happens: any factorization in K[x]K[x] of an R[x]R[x]-polynomial can be rescaled, coefficient by coefficient, into a genuine factorization in R[x]R[x], with the rescaling factors absorbed entirely into RR's own unique factorization of contents. Picture the set of nonzero polynomials in R[x]R[x] as fibered over the "shape" polynomials in K[x]K[x] (up to scaling), with each fiber being a copy of R∖{0}R \setminus \{0\} recording the content. The diagram below visualizes this fibration and how factorization respects it.

:::svg[Two layers of factorization: content in R, shape in K[x]] f = 6x² − 24 ∈ ℤ[x] content layer (in R = ℤ): c(f)=6 = 2 · 3 (unique in ℤ) shape layer (primitive part, factored in K[x]=ℚ[x]): x² − 4 = (x−2)(x+2) irreducible/ℚ each Gauss's Lemma: (x−2)(x+2) stays primitive in ℤ[x] — no hidden common integer factor appears in the product.

The central theorem

TheoremGauss's Lemma (multiplicativity of content)

Let RR be a UFD with fraction field KK. If f,g∈R[x]f, g \in R[x] are both primitive, then fgfg is primitive.

Proof

We argue by contradiction, using that RR is a UFD, hence has prime elements generating prime ideals (irreducibles equal primes in a UFD). Suppose fgfg is not primitive. Then some irreducible p∈Rp \in R divides every coefficient of fgfg. Since RR is a UFD, pp is prime, so the ideal (p)(p) is a prime ideal of RR, and the quotient R‾=R/(p)\overline{R} = R/(p) is an integral domain. Reduction mod pp gives a ring homomorphism

π:R[x]→R‾[x],π(∑aixi)=∑ai‾xi,\pi: R[x] \to \overline{R}[x], \qquad \pi\Big(\sum a_i x^i\Big) = \sum \overline{a_i} x^i,

which is a ring homomorphism because it is induced coefficientwise from the ring homomorphism R→R‾R \to \overline{R}. Because R‾\overline{R} is a domain, R‾[x]\overline{R}[x] is a domain (the leading coefficient of a product of nonzero polynomials over a domain is the product of leading coefficients, hence nonzero). Now p∣p \mid every coefficient of fgfg means exactly π(fg)=0\pi(fg) = 0 in R‾[x]\overline{R}[x]. Since π\pi is a ring homomorphism, π(fg)=π(f)π(g)\pi(fg) = \pi(f)\pi(g), so

π(f) π(g)=0 in R‾[x].\pi(f)\,\pi(g) = 0 \text{ in } \overline{R}[x].

As R‾[x]\overline{R}[x] is a domain, π(f)=0\pi(f) = 0 or π(g)=0\pi(g) = 0. Say π(f)=0\pi(f) = 0; this means every coefficient of ff reduces to 00 in R‾\overline{R}, i.e. pp divides every coefficient of ff. But then p∣c(f)p \mid c(f), contradicting that ff is primitive (as c(f)c(f) would fail to be a unit). This contradiction shows no irreducible pp divides every coefficient of fgfg, so fgfg is primitive. ■\blacksquare

∎
LemmaContent is multiplicative

Let RR be a UFD with fraction field KK, and let f,g∈R[x]f, g \in R[x] be nonzero. Then

c(fg)=c(f) c(g)(up to a unit of R).c(fg) = c(f)\,c(g) \quad \text{(up to a unit of } R\text{).}
Proof

Write f=c(f)f1f = c(f) f_1 and g=c(g)g1g = c(g) g_1 with f1,g1f_1, g_1 primitive. Then fg=c(f)c(g)f1g1fg = c(f)c(g) f_1 g_1. By Gauss's Lemma (multiplicativity of content) just proved, f1g1f_1 g_1 is primitive, so this expresses fgfg as (an element of RR) times (a primitive polynomial), and by definition of content this element of RR is c(fg)c(fg) up to a unit. Hence c(fg)=c(f)c(g)c(fg) = c(f)c(g) up to a unit. ■\blacksquare

∎

:::lemma[Primitive polynomial irreducible in K[x]K[x] stays irreducible in R[x]R[x]] Let RR be a UFD with fraction field KK. Let f∈R[x]f \in R[x] be primitive with deg⁡f≥1\deg f \ge 1. If ff is irreducible in K[x]K[x], then ff is irreducible in R[x]R[x].

Proof

Suppose f=ghf = gh with g,h∈R[x]g, h \in R[x] nonunits of R[x]R[x]. Since deg⁡f≥1\deg f \ge 1 and R[x]R[x]-units are exactly R×⊂RR^\times \subset R (units of R[x]R[x] have degree 00, as RR is a domain so degrees add), at least one of g,hg,h must have positive degree — in fact since ff irreducible in K[x]K[x] would be contradicted only by a genuine degree splitting, consider two cases. Case 1: both g,hg,h have positive degree. Viewing g,hg, h as elements of K[x]K[x] (via R⊂KR \subset K), this is a factorization of ff in K[x]K[x] into two factors of positive degree, hence neither is a unit of K[x]K[x] (units of K[x]K[x] are exactly K×K^\times, degree 00). This contradicts irreducibility of ff in K[x]K[x]. Case 2: one of them, say gg, has degree 00, so g∈Rg \in R, g≠0g \neq 0. Then gg divides every coefficient of f=ghf = gh, so g∣c(f)g \mid c(f) in RR. Since ff is primitive, c(f)c(f) is a unit, so gg is a unit of RR (a divisor of a unit is a unit, as RR is a domain: if gd=c(f)=:ug d = c(f) =: u with uu a unit, then g(du−1)=1g(du^{-1}) = 1). This makes gg a unit of R[x]R[x], contradicting that gg was assumed a nonunit. Both cases give a contradiction, so no such factorization f=ghf = gh with both nonunits exists: ff is irreducible in R[x]R[x]. ■\blacksquare

∎

:::theorem[R UFD implies R[x] UFD] Let RR be a unique factorization domain. Then R[x]R[x] is a unique factorization domain. The irreducible elements of R[x]R[x] are exactly: (i) the irreducible elements of RR (viewed as degree-00 polynomials), and (ii) the primitive polynomials f∈R[x]f \in R[x] of degree ≥1\ge 1 that are irreducible in K[x]K[x], where K=Frac(R)K = \mathrm{Frac}(R).

Proof

We must show (A) existence of factorization into irreducibles for every nonzero nonunit of R[x]R[x], and (B) uniqueness of such a factorization up to order and units. We use the fact, taken as known, that K[x]K[x] is a Euclidean domain (division algorithm by degree), hence a PID, hence a UFD — this is a standard fact about polynomial rings over a field and does not depend on the theorem being proved here.

Step 0: identify units and irreducibles as claimed. Since RR is a domain, for f,g∈R[x]f,g \in R[x] nonzero, deg⁡(fg)=deg⁡f+deg⁡g\deg(fg) = \deg f + \deg g (leading coefficients multiply to something nonzero, as RR is a domain). Hence a unit of R[x]R[x] must have degree 00 and be a unit of RR; conversely units of RR are units of R[x]R[x]. So R[x]×=R×R[x]^\times = R^\times. A degree-00 element p∈R⊂R[x]p \in R \subset R[x] is irreducible in R[x]R[x] iff it is irreducible in RR: any factorization of pp in R[x]R[x] forces both factors to have degree 00 (degrees are nonnegative and sum to 00), hence lie in RR, reducing to factorization in RR; the converse is immediate. A polynomial ff of degree ≥1\ge 1 that is irreducible in R[x]R[x] must be primitive: if c(f)c(f) were a nonunit, c(f)c(f) would be a nonunit factor of RR times a nonunit (degree ≥1\ge 1 implies nonunit) primitive part, giving a nontrivial factorization, contradicting irreducibility. It must also be irreducible in K[x]K[x]: this is proved as the (easy) converse direction of [[Primitive polynomial irreducible in K[x] stays irreducible in R[x]]] — if ff factored nontrivially in K[x]K[x] as f=GHf = GH with deg⁡G,deg⁡H≥1\deg G, \deg H \ge 1 (a factor of degree 00 in K[x]K[x] would be a unit, and if ff had a factorization with one factor of degree 00 that is a nonunit of KK this is vacuous since KK is a field, all nonzero constants are units), then clearing denominators via contents (as in Step 1 below) yields a nontrivial R[x]R[x]-factorization of ff, contradicting irreducibility in R[x]R[x]. Conversely [[Primitive polynomial irreducible in K[x] stays irreducible in R[x]]] shows every primitive, K[x]K[x]-irreducible polynomial of degree ≥1\ge 1 is irreducible in R[x]R[x]. This confirms the list (i), (ii) is exactly the set of irreducibles of R[x]R[x].

Step 1 (Existence). Let f∈R[x]f \in R[x] be a nonzero nonunit. Write f=c(f)f1f = c(f) f_1 with f1f_1 primitive. If deg⁡f=0\deg f = 0, then f1f_1 is a unit (primitive of degree 00 means c(f1)=f1c(f_1)=f_1 is a unit, since a degree 0 primitive polynomial is a unit of RR by definition of primitivity: content of f1f_1 is f1f_1 itself which must be a unit), so f=c(f)⋅(unit)f = c(f) \cdot(\text{unit}), i.e. ff is (up to unit) equal to c(f)∈Rc(f) \in R, a nonzero nonunit of RR since ff is a nonunit of R[x]R[x] and has degree 00. As RR is a UFD, c(f)c(f) factors into irreducibles of RR, giving a factorization of ff into irreducibles of type (i). If deg⁡f≥1\deg f \ge 1: first factor c(f)c(f) in RR into irreducibles of RR (possibly c(f)c(f) is a unit, contributing no factors) — this uses that RR is a UFD. Next, factor f1f_1 (primitive, degree ≥1\ge 1) into irreducibles in K[x]K[x], using that K[x]K[x] is a PID hence UFD:

f1=G1G2⋯Gr,Gi∈K[x] irreducible, deg⁡Gi≥1.f_1 = G_1 G_2 \cdots G_r, \qquad G_i \in K[x] \text{ irreducible}, \ \deg G_i \ge 1.

Each GiG_i has coefficients in KK; clear denominators by choosing di∈R∖{0}d_i \in R\setminus\{0\} so that diGi∈R[x]d_i G_i \in R[x] (multiply through by a common denominator of the coefficients of GiG_i, which exists since K=Frac(R)K = \mathrm{Frac}(R)). Write diGi=cigid_i G_i = c_i g_i with gi∈R[x]g_i \in R[x] primitive and ci∈Rc_i \in R (extracting content). Then gig_i is a primitive polynomial in R[x]R[x], associate in K[x]K[x] to GiG_i (they differ by the scalar ci/di∈K×c_i/d_i \in K^\times), hence gig_i is also irreducible in K[x]K[x] (irreducibility is preserved under multiplying by a unit of K[x]K[x], i.e. a nonzero scalar). By [[Primitive polynomial irreducible in K[x] stays irreducible in R[x]]], each gig_i is irreducible in R[x]R[x]. Now

(∏idi)f1=(∏idi)∏iGi=∏i(diGi)=∏i(cigi)=(∏ici)∏igi.\Big(\prod_i d_i\Big) f_1 = \Big(\prod_i d_i\Big) \prod_i G_i = \prod_i (d_i G_i) = \prod_i (c_i g_i) = \Big(\prod_i c_i\Big)\prod_i g_i.

Taking contents of both sides (multiplicative by the Lemma Content is multiplicative, and c(f1)c(f_1) is a unit since f1f_1 primitive, and c(gi)c(g_i) is a unit for each ii since each gig_i is primitive), we get

∏idi ∼ ∏ici(associates in R),\prod_i d_i \ \sim\ \prod_i c_i \quad \text{(associates in } R),

where ∼\sim denotes equality up to a unit of RR. So ∏di=u∏ci\prod d_i = u \prod c_i for a unit u∈R×u \in R^\times. Substituting back:

f1=u∏igi.f_1 = u \prod_i g_i.

Thus f1f_1 factors into irreducibles gig_i of R[x]R[x] (absorbing the unit uu into g1g_1, say). Combining with the factorization of c(f)c(f) into irreducibles of RR, we obtain

f=c(f)⋅f1=(irreducibles of R)×(irreducibles gi of R[x]),f = c(f) \cdot f_1 = (\text{irreducibles of } R) \times (\text{irreducibles } g_i \text{ of } R[x]),

a full factorization of ff into irreducibles of R[x]R[x].

Step 2 (Uniqueness). Suppose

f=p1⋯ps⋅g1⋯gr=q1⋯qt⋅h1⋯hmf = p_1 \cdots p_s \cdot g_1 \cdots g_r = q_1 \cdots q_t \cdot h_1 \cdots h_m

are two factorizations of ff into irreducibles of R[x]R[x], where pi,qj∈Rp_i, q_j \in R are the type-(i) (degree 00) irreducible factors and gk,hlg_k, h_l are the type-(ii) (primitive, positive-degree) irreducible factors, listed separately. Taking contents of both sides and using multiplicativity of content (Content is multiplicative) plus the fact that each gk,hlg_k, h_l is primitive (content a unit) and each pi,qj∈Rp_i, q_j \in R has content itself:

c(f)=p1⋯ps⋅u1=q1⋯qt⋅u2c(f) = p_1 \cdots p_s \cdot u_1 = q_1 \cdots q_t \cdot u_2

for units u1,u2∈R×u_1, u_2 \in R^\times absorbing the unit contents of the primitive parts. So p1⋯psp_1 \cdots p_s and q1⋯qtq_1 \cdots q_t are associate factorizations of c(f)∈Rc(f) \in R into irreducibles of RR (up to reordering and units). Since RR is a UFD, s=ts = t and after reindexing qi=wipiq_i = w_i p_i for units wi∈R×w_i \in R^\times. Next, dividing ff by c(f)c(f) (working in K[x]K[x], where division by any nonzero constant is valid since KK is a field), we get that

g1⋯grandh1⋯hmg_1 \cdots g_r \quad \text{and} \quad h_1 \cdots h_m

are both, up to units of RR, equal to the primitive part f1f_1 of ff, hence they are equal up to a unit of K×K^\times as elements of K[x]K[x]:

g1⋯gr=λ h1⋯hm,λ∈K×.g_1 \cdots g_r = \lambda\, h_1 \cdots h_m, \qquad \lambda \in K^\times.

Each gk,hlg_k, h_l is irreducible in K[x]K[x] (this was shown in Step 0 to be automatic for type-(ii) irreducibles of R[x]R[x]). Since K[x]K[x] is a UFD (it is a PID), unique factorization in K[x]K[x] gives r=mr = m and, after reindexing, hl=μlglh_l = \mu_l g_l for some μl∈K×\mu_l \in K^\times. It remains to see each μl\mu_l is actually a unit of RR, not merely of KK. Since glg_l and hlh_l are both primitive in R[x]R[x] (by hypothesis, as type-(ii) irreducibles), write μl=a/b\mu_l = a/b in lowest terms with a,b∈Ra, b \in R, gcd⁡(a,b)\gcd(a,b) a unit (using unique factorization in RR to reduce to lowest terms). Then b hl=a glb\, h_l = a\, g_l in R[x]R[x]. Taking contents: c(bhl)=b c(hl)∼bc(b h_l) = b\, c(h_l) \sim b (since hlh_l primitive) and c(agl)=a c(gl)∼ac(a g_l) = a\, c(g_l) \sim a (since glg_l primitive), so b∼ab \sim a in RR, i.e. a=ϵba = \epsilon b for a unit ϵ∈R×\epsilon \in R^\times. Since gcd⁡(a,b)\gcd(a,b) is a unit and a=ϵba = \epsilon b, we must have bb itself a unit (else the prime factors of bb would divide aa too, contradicting the gcd being a unit — unless bb has no prime factors, i.e. bb is a unit). Hence μl=a/b=ϵ∈R×\mu_l = a/b = \epsilon \in R^\times is a unit of RR. So hl=ϵlglh_l = \epsilon_l g_l for units ϵl∈R×\epsilon_l \in R^\times, for each ll. This, together with qi=wipiq_i = w_i p_i, establishes that the two factorizations of ff agree up to reordering and units of R[x]R[x] (recall R[x]×=R×R[x]^\times = R^\times). This proves uniqueness, completing the proof that R[x]R[x] is a UFD. ■\blacksquare

∎
CorollaryPolynomial rings in several variables over a UFD

If RR is a UFD, then R[x1,…,xn]R[x_1, \dots, x_n] is a UFD for every n≥1n \ge 1.

Proof

Induct on nn. The base case n=1n=1 is the theorem just proved. For the inductive step, note R[x1,…,xn]≅R[x1,…,xn−1][xn]R[x_1,\dots,x_n] \cong R[x_1,\dots,x_{n-1}][x_n] as rings (a standard and routine isomorphism: both sides consist of polynomials in xnx_n with coefficients that are polynomials in x1,…,xn−1x_1,\dots,x_{n-1}, with the evident matching ring operations). By the inductive hypothesis R[x1,…,xn−1]R[x_1,\dots,x_{n-1}] is a UFD, so applying the theorem again (with RR replaced by R[x1,…,xn−1]R[x_1,\dots,x_{n-1}]) shows R[x1,…,xn−1][xn]R[x_1,\dots,x_{n-1}][x_n], hence R[x1,…,xn]R[x_1,\dots,x_n], is a UFD. ■\blacksquare

∎

Worked examples

Example 1 — factoring an asymmetric integer polynomial in Z[x]\mathbb{Z}[x]. Let R=ZR = \mathbb{Z} (a UFD, indeed a PID), K=QK = \mathbb{Q}, and consider

f(x)=12x4−3x3−54x2+27x+90∈Z[x].f(x) = 12x^4 - 3x^3 - 54x^2 + 27x + 90 \in \mathbb{Z}[x].

We factor ff completely into irreducibles of Z[x]\mathbb{Z}[x], using the content/primitive-part decomposition central to the theorem. Step 1: extract the content. The coefficients are 12,−3,−54,27,9012, -3, -54, 27, 90. Their gcd: 12=22⋅312 = 2^2\cdot 3, 3=33 = 3, 54=2⋅3354 = 2\cdot 3^3, 27=3327=3^3, 90=2⋅32⋅590 = 2\cdot 3^2\cdot 5. The common factor is 33, so c(f)=3c(f) = 3 and

f=3(4x4−x3−18x2+9x+30)=3f1(x),f = 3\big(4x^4 - x^3 - 18x^2 + 9x + 30\big) = 3 f_1(x),

where f1f_1 has coefficients 4,−1,−18,9,304,-1,-18,9,30 with gcd 11: indeed gcd⁡(4,−1)=1\gcd(4,-1)=1 already, so f1f_1 is primitive. Step 2: factor f1f_1 over Q\mathbb{Q} (equivalently, search for rational roots and quadratic factors). Try the rational root test on f1f_1: candidate roots p/qp/q with p∣30p \mid 30, q∣4q \mid 4. Testing x=−2x = -2: 4(16)−(−8)−18(4)+9(−2)+30=64+8−72−18+30=12≠04(16) - (-8) - 18(4) + 9(-2) + 30 = 64 + 8 - 72 - 18 + 30 = 12 \ne 0. Testing x=5/2x = 5/2: 4(625/16)−(125/8)−18(25/4)+9(5/2)+304(625/16) - (125/8) - 18(25/4) + 9(5/2) + 30. Compute: 4⋅625/16=625/44\cdot 625/16 = 625/4, minus 125/8125/8, minus 450/4450/4, plus 45/245/2, plus 3030. Common denominator 88: 1250/8−125/8−900/8+180/8+240/8=(1250−125−900+180+240)/8=645/8≠01250/8 - 125/8 - 900/8 + 180/8 + 240/8 = (1250-125-900+180+240)/8 = 645/8 \ne 0. Instead factor by grouping, guided by the structure 4x4−18x2+304x^4 - 18x^2 + 30 and −x3+9x-x^3+9x:

f1=4x4−x3−18x2+9x+30.f_1 = 4x^4 - x^3 - 18x^2 + 9x + 30.

Group as (4x4−18x2+30)−(x3−9x)=2(2x4−9x2+15)−x(x2−9)(4x^4 - 18x^2 + 30) - (x^3 - 9x) = 2(2x^4 - 9x^2+15) - x(x^2-9). This does not factor cleanly, so instead try a direct quadratic-times-quadratic ansatz with integer coefficients (valid to try in Z[x]\mathbb{Z}[x] directly since we expect an integral factorization, and if one exists over Q\mathbb{Q} Gauss's Lemma guarantees an integral one for a primitive polynomial):

f1=(2x2+ax+b)(2x2+cx+d),a,b,c,d∈Z.f_1 = (2x^2 + ax + b)(2x^2 + cx + d), \quad a,b,c,d \in \mathbb{Z}.

Expanding: 4x4+2(a+c)x3+(ac+2b+2d)x2+(ad+bc)x+bd4x^4 + 2(a+c)x^3 + (ac + 2b+2d)x^2 + (ad+bc)x + bd. Matching to 4x4−x3−18x2+9x+304x^4 -x^3 -18x^2+9x+30:

2(a+c)=−1.2(a+c) = -1.

This has no integer solution! So a 2-22\text{-}2 split with leading coefficients 2,22,2 fails over Z\mathbb{Z}. Try leading coefficients 4,14,1:

f1=(4x2+ax+b)(x2+cx+d),a,b,c,d∈Z.f_1 = (4x^2+ax+b)(x^2+cx+d), \quad a,b,c,d\in\mathbb{Z}.

Expand: 4x4+(4c+a)x3+(4d+ac+b)x2+(ad+bc)x+bd4x^4 + (4c+a)x^3 + (4d+ac+b)x^2+(ad+bc)x+bd. Match:

4c+a=−1,4d+ac+b=−18,ad+bc=9,bd=30.4c+a=-1,\quad 4d+ac+b=-18,\quad ad+bc=9,\quad bd=30.

Try bd=30bd=30 with b=−3,d=−10b=-3,d=-10: then ad+bc=−10a−3c=9ad+bc = -10a-3c=9 and 4c+a=−1⇒a=−1−4c4c+a=-1 \Rightarrow a=-1-4c. Substitute: −10(−1−4c)−3c=10+40c−3c=10+37c=9⇒37c=−1-10(-1-4c)-3c = 10+40c-3c=10+37c=9 \Rightarrow 37c=-1, not integral. Try b=10,d=3b=10, d=3: bd=30bd=30. Then ad+bc=3a+10c=9ad+bc=3a+10c=9, and a=−1−4ca=-1-4c, so 3(−1−4c)+10c=−3−12c+10c=−3−2c=9⇒−2c=12⇒c=−63(-1-4c)+10c = -3-12c+10c=-3-2c=9\Rightarrow -2c=12\Rightarrow c=-6, then a=−1−4(−6)=23a=-1-4(-6)=23. Check third equation: 4d+ac+b=12+23(−6)+10=12−138+10=−116≠−184d+ac+b = 12 + 23(-6)+10 = 12-138+10=-116\ne -18. Fails. Try b=−10,d=−3b=-10,d=-3: ad+bc=−3a−10c=9ad+bc = -3a-10c=9, a=−1−4ca=-1-4c, so −3(−1−4c)−10c=3+12c−10c=3+2c=9⇒c=3-3(-1-4c)-10c=3+12c-10c=3+2c=9\Rightarrow c=3, a=−1−12=−13a=-1-12=-13. Check: 4d+ac+b=−12+(−13)(3)+(−10)=−12−39−10=−61≠−184d+ac+b=-12+(-13)(3)+(-10)=-12-39-10=-61\ne-18. Fails. Try b=6,d=5b=6,d=5: ad+bc=5a+6c=9ad+bc=5a+6c=9, a=−1−4ca=-1-4c: 5(−1−4c)+6c=−5−20c+6c=−5−14c=9⇒−14c=14⇒c=−15(-1-4c)+6c=-5-20c+6c=-5-14c=9\Rightarrow -14c=14\Rightarrow c=-1, a=−1+4=3a=-1+4=3. Check: 4d+ac+b=20+3(−1)+6=20−3+6=23≠−184d+ac+b=20+3(-1)+6=20-3+6=23\ne-18. Fails. Try b=−6,d=−5b=-6,d=-5: 5a⋅(−1)5a\cdot(-1)... let's redo generally: ad+bc=−5a−6c=9ad+bc=-5a-6c=9, a=−1−4ca=-1-4c: −5(−1−4c)−6c=5+20c−6c=5+14c=9⇒c=2/7-5(-1-4c)-6c=5+20c-6c=5+14c=9\Rightarrow c=2/7. Not integral. Try b=15,d=2b=15,d=2: ad+bc=2a+15c=9ad+bc=2a+15c=9; a=−1−4ca=-1-4c: 2(−1−4c)+15c=−2−8c+15c=−2+7c=9⇒c=11/72(-1-4c)+15c=-2-8c+15c=-2+7c=9\Rightarrow c=11/7. Fail. Try b=−15,d=−2b=-15,d=-2: 2a⋅(−1)2a\cdot(-1): ad+bc=−2a−15c=9ad+bc=-2a-15c=9; a=−1−4ca=-1-4c: −2(−1−4c)−15c=2+8c−15c=2−7c=9⇒c=−1-2(-1-4c)-15c=2+8c-15c=2-7c=9\Rightarrow c=-1, a=−1+4=3a=-1+4=3. Check: 4d+ac+b=−8+3(−1)+(−15)=−8−3−15=−26≠−184d+ac+b=-8+3(-1)+(-15)=-8-3-15=-26\ne-18. Fail. Try b=2,d=15b=2,d=15: ad+bc=15a+2c=9ad+bc=15a+2c=9; a=−1−4ca=-1-4c: 15(−1−4c)+2c=−15−60c+2c=−15−58c=9⇒c=−24/5815(-1-4c)+2c=-15-60c+2c=-15-58c=9\Rightarrow c=-24/58, no. Try b=−2,d=−15b=-2,d=-15: ad+bc=−15a−2c=9ad+bc=-15a-2c=9; a=−1−4ca=-1-4c: −15(−1−4c)−2c=15+60c−2c=15+58c=9⇒c=−6/58-15(-1-4c)-2c=15+60c-2c=15+58c=9\Rightarrow c=-6/58, no. Try b=30,d=1b=30,d=1: ad+bc=a+30c=9ad+bc=a+30c=9; a=−1−4ca=-1-4c: −1−4c+30c=−1+26c=9⇒c=10/26-1-4c+30c=-1+26c=9\Rightarrow c=10/26, no. Try b=−30,d=−1b=-30,d=-1: ad+bc=−a−30c=9ad+bc=-a-30c=9; a=−1−4ca=-1-4c: 1+4c−30c=1−26c=9⇒c=−8/261+4c-30c=1-26c=9\Rightarrow c=-8/26, no. Try b=1,d=30b=1,d=30: ad+bc=30a+c=9ad+bc=30a+c=9; a=−1−4ca=-1-4c: 30(−1−4c)+c=−30−120c+c=−30−119c=9⇒c=−39/11930(-1-4c)+c=-30-120c+c=-30-119c=9\Rightarrow c=-39/119, no. Try b=−1,d=−30b=-1,d=-30: ad+bc=−30a−c=9ad+bc=-30a-c=9; a=−1−4ca=-1-4c: 30+120c−c=30+119c=9⇒c=−21/11930+120c-c=30+119c=9\Rightarrow c=-21/119, no. All integer factorizations of 3030 into b,db,d (with matching signs, since bd=30>0bd=30>0) exhausted (up to swapping b↔db\leftrightarrow d, which just swaps the two quadratic factors, already covered): no 4-14\text{-}1 quadratic split works. Combined with the failed 2-22\text{-}2 split, f1f_1 has no factorization into two quadratics over Z\mathbb{Z}, hence (by Gauss's Lemma (multiplicativity of content) applied in reverse — a factorization over Q\mathbb{Q} into a degree 22 times degree 22 would rescale to an integral one) none over Q\mathbb{Q} either. Having ruled out quadratic-quadratic and (implicitly, since no rational root was found among reasonable candidates) linear-cubic splits, we check remaining rational root candidates more systematically. Since f1(1)=4−1−18+9+30=24f_1(1)=4-1-18+9+30=24, f1(−1)=4+1−18−9+30=8f_1(-1)=4+1-18-9+30=8, f1(2)=64−8−72+18+30=32f_1(2)=64-8-72+18+30=32, f1(−2)=64+8−72−18+30=12f_1(-2)=64+8-72-18+30=12, f1(3)=324−27−162+27+30=192f_1(3)=324-27-162+27+30=192, f1(−3)=324+27−162−27+30=192f_1(-3)=324+27-162-27+30=192, f1(3/2)=4(81/16)−27/8−18(9/4)+27/2+30=81/4−27/8−81/2+27/2+30f_1(3/2)=4(81/16)-27/8-18(9/4)+27/2+30 = 81/4-27/8-81/2+27/2+30; common denom 8: 162/8−27/8−324/8+108/8+240/8=159/8≠0162/8-27/8-324/8+108/8+240/8=159/8\ne0; f1(−5/2)f_1(-5/2): computed pattern above nonzero; f1(−3/2)f_1(-3/2): 4(81/16)+27/8−18(9/4)−27/2+30=81/4+27/8−81/2−27/2+304(81/16)+27/8-18(9/4)-27/2+30=81/4+27/8-81/2-27/2+30, denom 8: 162/8+27/8−324/8−108/8+240/8=−3/8≠0162/8+27/8-324/8-108/8+240/8=-3/8\ne0; f1(−1/2)f_1(-1/2): 4(1/16)+1/8−18/4−9/2+30=1/4+1/8−9/2−9/2+304(1/16)+1/8-18/4-9/2+30=1/4+1/8-9/2-9/2+30, denom 8: 2/8+1/8−72/8−72/8+240/8=99/8≠02/8+1/8-72/8-72/8+240/8=99/8\ne0; f1(1/2)f_1(1/2): 1/4−1/8−9/2+9/2+30=1/4−1/8+30=1/8+30=241/8≠01/4-1/8-9/2+9/2+30 = 1/4-1/8+30 = 1/8+30 = 241/8\ne 0. No rational roots found among the divisors tested, and combined with the exhaustive quadratic search, f1f_1 is irreducible over Q\mathbb{Q}, hence irreducible in Z[x]\mathbb{Z}[x] by [[Primitive polynomial irreducible in K[x] stays irreducible in R[x]]] (it is primitive by Step 1). Conclusion. The complete factorization of ff into irreducibles of Z[x]\mathbb{Z}[x] is

f(x)=3⋅(4x4−x3−18x2+9x+30),f(x) = 3 \cdot \big(4x^4 - x^3 - 18x^2 + 9x + 30\big),

with 33 irreducible in Z\mathbb{Z} (type (i)) and the quartic irreducible in Z[x]\mathbb{Z}[x] as a primitive, Q\mathbb{Q}-irreducible polynomial (type (ii)). This factorization is unique up to reordering and sign, exactly as the theorem guarantees.

Example 2 — a two-variable factorization in k[x,y]k[x,y] illustrating the induction step. Let R=k[y]R = k[y] for a field kk (a UFD, in fact a PID, since k[y]k[y] is Euclidean), and consider R[x]=k[y][x]≅k[x,y]R[x] = k[y][x] \cong k[x,y]. Factor

f(x,y)=2y(y−1) x2  −  2y(y−1)(2y+3) x  +  2y(y−1)(y+1)(y+2) ∈k[x,y],f(x,y) = 2y(y-1)\,x^2 \;-\; 2y(y-1)(2y+3)\,x \;+\; 2y(y-1)(y+1)(y+2) \ \in k[x,y],

assuming char(k)≠2\mathrm{char}(k) \ne 2 so that 2∈k[y]2 \in k[y] is a unit (to keep the content computation transparent, take k=Qk = \mathbb{Q} for concreteness, so 2∈R×2 \in R^\times is irrelevant to content and we may instead track the genuinely non-unit factor). Step 1: extract the content in R=k[y]R = k[y]. Viewing ff as a polynomial in xx with coefficients in k[y]k[y]:

a2=2y(y−1),a1=−2y(y−1)(2y+3),a0=2y(y−1)(y+1)(y+2).a_2 = 2y(y-1), \qquad a_1 = -2y(y-1)(2y+3), \qquad a_0 = 2y(y-1)(y+1)(y+2).

Every coefficient is divisible by 2y(y−1)2y(y-1) (using 2∈k×⊂R×2 \in k^\times \subset R^\times, so up to a unit the common factor is y(y−1)y(y-1)), and no further common irreducible factor of k[y]k[y] divides all three: a2/(2y(y−1))=1a_2/(2y(y-1)) = 1 is a unit, so the gcd of the three "quotient" coefficients 1,−(2y+3),(y+1)(y+2)1, -(2y+3), (y+1)(y+2) is a unit (since one of them, 11, is already a unit). Hence

c(f)=y(y−1)(up to the unit 2∈k×),c(f) = y(y-1) \quad (\text{up to the unit } 2 \in k^\times),

and

f=2y(y−1)[x2−(2y+3)x+(y+1)(y+2)]=c(f)⋅f1(x,y),f = 2y(y-1)\big[x^2 - (2y+3)x + (y+1)(y+2)\big] = c(f)\cdot f_1(x,y),

where f1=x2−(2y+3)x+(y+1)(y+2)f_1 = x^2 - (2y+3)x+(y+1)(y+2) is primitive in R[x]R[x] (its coefficient 11 on x2x^2 shows the content is a unit). Step 2: factor f1f_1 in K[x]K[x] where K=k(y)=Frac(k[y])K = k(y) = \mathrm{Frac}(k[y]). Solve the quadratic in xx with coefficients in k(y)k(y): discriminant

Δ=(2y+3)2−4(y+1)(y+2)=4y2+12y+9−4(y2+3y+2)=4y2+12y+9−4y2−12y−8=1.\Delta = (2y+3)^2 - 4(y+1)(y+2) = 4y^2+12y+9 - 4(y^2+3y+2) = 4y^2+12y+9-4y^2-12y-8 = 1.

So x=(2y+3)±12x = \dfrac{(2y+3) \pm 1}{2}, giving roots x=y+2x = y+2 and x=y+1x = y+1. Thus

f1=(x−(y+1))(x−(y+2))in k(y)[x],f_1 = (x-(y+1))(x-(y+2)) \quad \text{in } k(y)[x],

and in fact both factors already lie in k[y][x]=R[x]k[y][x] = R[x] with coefficients in RR — no denominators needed to clear, consistent with Gauss's Lemma (the factorization over the fraction field was already integral). Step 3: verify primitivity and irreducibility of each linear factor. Each of x−(y+1)x-(y+1) and x−(y+2)x-(y+2) is primitive in R[x]=k[y][x]R[x] = k[y][x]: its content (gcd of coefficients 11 and −(y+1)-(y+1), resp. 11 and −(y+2)-(y+2)) is 11, a unit, since the leading coefficient is already 11. Each has degree 1≥11 \ge 1 in xx, hence is irreducible in K[x]K[x] (degree 11 polynomials over a field are always irreducible), so by [[Primitive polynomial irreducible in K[x] stays irreducible in R[x]]] each is irreducible in R[x]=k[y][x]R[x] = k[y][x]. Step 4: factor the content c(f)=2y(y−1)c(f) = 2y(y-1) in R=k[y]R = k[y]. Here 2∈k×2 \in k^\times is a unit, yy is irreducible in k[y]k[y] (degree 11), and y−1y - 1 is irreducible in k[y]k[y] (degree 11). So c(f)=2⋅y⋅(y−1)c(f) = 2\cdot y \cdot (y-1) is the factorization into irreducibles of RR (with the unit 22 absorbed). Conclusion. The complete factorization of ff into irreducibles of k[x,y]≅k[y][x]k[x,y] \cong k[y][x] is

f(x,y)=2⋅y⋅(y−1)⋅(x−(y+1))⋅(x−(y+2)),f(x,y) = 2\cdot y\cdot (y-1)\cdot \big(x-(y+1)\big)\cdot\big(x-(y+2)\big),

exhibiting exactly the theorem's structure: irreducibles of R=k[y]R=k[y] (namely yy and y−1y-1, up to the unit 22) times primitive, KK-irreducible polynomials of positive xx-degree (the two linear factors in xx).

:::counterexample[Dropping "R is a UFD": failure of unique factorization in R[x]] The hypothesis that RR itself is a UFD cannot be dropped: if RR is merely an integral domain without unique factorization, R[x]R[x] can fail to be a UFD even though one might hope the polynomial variable "adds room" to factor things uniquely. Let R=Z[−5]={a+b−5:a,b∈Z}R = \mathbb{Z}[\sqrt{-5}] = \{a+b\sqrt{-5} : a,b \in \mathbb{Z}\}, a classical integral domain that is not a UFD, since

6=2⋅3=(1+−5)(1−−5)6 = 2 \cdot 3 = (1+\sqrt{-5})(1-\sqrt{-5})

are two genuinely different factorizations of 66 into irreducibles of RR (none of 2,3,1±−52,3,1\pm\sqrt{-5} is a unit or associate of another; this is the standard example showing Z[−5]\mathbb{Z}[\sqrt{-5}] is not a UFD, verified via the norm N(a+b−5)=a2+5b2N(a+b\sqrt{-5})=a^2+5b^2: N(2)=4N(2)=4, N(3)=9N(3)=9, N(1±−5)=6N(1\pm\sqrt{-5})=6, and no elements of norm 22 or 33 exist since a2+5b2=2a^2+5b^2=2 or 33 has no integer solutions). Now consider the constant polynomial 6∈R[x]6 \in R[x] (degree 00). It is a nonzero nonunit of R[x]R[x] (since R[x]×=R×={±1}R[x]^\times = R^\times = \{\pm 1\}, as one checks directly that R×={1,−1}R^\times=\{1,-1\} via the norm being multiplicative and equal to 11 only for units). It has (at least) the two factorizations

6=2⋅3=(1+−5)(1−−5)6 = 2\cdot 3 = (1+\sqrt{-5})(1-\sqrt{-5})

already inside R[x]R[x] (these are degree-00 elements, so the factorization is literally the same one as in RR, now viewed inside the larger ring R[x]R[x]). Each of 2,3,1+−5,1−−52, 3, 1+\sqrt{-5}, 1-\sqrt{-5} remains irreducible in R[x]R[x]: a degree-00 element p∈R⊂R[x]p \in R \subset R[x] is irreducible in R[x]R[x] if and only if it is irreducible in RR (exactly as shown in Step 0 of the proof above — the argument used only that RR is a domain, not that RR is a UFD, so it applies here too), and each of these four elements is irreducible in RR by the norm computation above. Since 2⋅32\cdot 3 and (1+−5)(1−−5)(1+\sqrt{-5})(1-\sqrt{-5}) are inequivalent factorizations of 66 into irreducibles — no irreducible on one side is an associate of an irreducible on the other, because their norms (4,94,9 versus 6,66,6) do not match up under multiplication by the only units ±1\pm1 — R[x]R[x] fails unique factorization for the element 66. Hence R[x]R[x] is not a UFD. This shows precisely why the hypothesis "RR is a UFD" in the theorem is essential: without it, even constant polynomials already violate unique factorization, before polynomials of positive degree are even considered.

Scope, edge cases & common traps

SituationWhat happensWhy
RR a fieldR[x]R[x] is a PID (stronger than UFD)division algorithm available directly
RR a PID but not a field (e.g. Z\mathbb{Z})R[x]R[x] is a UFD but not a PIDe.g. (2,x)⊂Z[x](2,x) \subset \mathbb{Z}[x] is not principal
RR a UFD, several variablesR[x1,…,xn]R[x_1,\dots,x_n] is a UFD for all nninduction on the tower R[x1]⋯[xn]R[x_1]\cdots[x_n]
RR not a UFDR[x]R[x] can fail to be a UFD, already at degree 00see counterexample above
Confusing "irreducible in K[x]K[x]" with "irreducible in R[x]R[x]"a nonprimitive polynomial can be K[x]K[x]-irreducible-looking but factor via its content in R[x]R[x]always separate content and primitive part first

The mathematician behind it

Origins · 1 minCarl Friedrich Gauss and how this was discovered
CF
Carl Friedrich Gauss
1777–1855 · Göttingen, foundational work in number theory and algebra

Gauss proved the multiplicativity of content — now called Gauss's Lemma — in his 1801 treatise Disquisitiones Arithmeticae, in the course of studying integer polynomials and the arithmetic of quadratic forms. He was concerned with showing that factorizations of integer polynomials could not "hide" behind rational coefficients, a question central to his broader program of putting the theory of forms and congruences on rigorous algebraic footing. The abstract UFD formulation used today is a later generalization, but the core mechanism — reducing modulo a prime to detect hidden common factors — is exactly Gauss's original argument.

History
- 1801 — Gauss publishes Disquisitiones Arithmeticae, proving what is now called Gauss's Lemma for Z[x]\mathbb{Z}[x].
- 1882 — Kronecker and Dedekind, developing ideal theory, generalize factorization arguments to broader classes of rings.
- Early 1900s — the abstract notion of unique factorization domain crystallizes within the emerging framework of commutative ring theory.
- 1920s–1930s — Emmy Noether and B.L. van der Waerden formalize the general UFD axioms and the transfer theorem for R[x]R[x] in modern algebraic language.
- Mid 20th century — the result becomes a standard tool in algebraic geometry, via coordinate rings k[x1,…,xn]k[x_1,\dots,x_n] of affine space.
Does the theorem require R to be Noetherian or have any finiteness condition?

No finiteness hypothesis is needed beyond RR being a UFD. The proof only uses: existence and uniqueness of factorizations in RR, the notion of content and its multiplicativity (Gauss's Lemma), and the fact that K[x]K[x] is a PID for K=Frac(R)K = \mathrm{Frac}(R) a field. UFDs need not be Noetherian in general (e.g. a polynomial ring in infinitely many variables over a field is a UFD but not Noetherian), and the theorem applies to those cases as well.

How does this connect to Eisenstein's Criterion?

Eisenstein's Criterion is a sufficient condition for a polynomial in R[x]R[x] (typically R=ZR=\mathbb{Z}) to be irreducible in K[x]K[x], phrased via a prime pp dividing all but the leading coefficient and p2p^2 not dividing the constant term. The proof of Eisenstein's Criterion uses the same reduction-mod-pp technique as the proof of Gauss's Lemma here, and its conclusion — irreducibility over K[x]K[x] — combines with this article's [[Primitive polynomial irreducible in K[x] stays irreducible in R[x]]] to yield irreducibility directly in R[x]R[x], provided the polynomial is primitive.

Can content and primitive part fail to be well-defined?

They are well-defined only up to multiplication by a unit of RR, exactly like a gcd in any UFD is only defined up to units. Any two valid choices of content differ by a unit, and correspondingly the primitive parts differ by the inverse unit. This ambiguity is harmless for the theorem, since "irreducible up to a unit" and "unique up to units" are exactly the equivalence built into the definition of unique factorization.

Before this