Gauss's Lemma

Gauss's Lemma underlies the fact that $R[x]$ is a UFD whenever $R$ is, which is the foundation for polynomial factorization algorithms used in computer algebra systems (Maple, Mathematica, Sage) when factoring integer or multivariate polynomials. It is the essential lemma behind Eisenstein's Criterion, one of the most widely used practical irreducibility tests, applied constantly in constructing field extensions and proving polynomials irreducible over $\mathbb{Q}$. In algebraic number theory, content arguments appear when studying minimal polynomials of algebraic integers and when clearing denominators in rings of integers. In cryptography, irreducibility testing over finite fields and over $\mathbb{Z}$ (e.g. for constructing finite fields $\mathbb{F}_{p^n}$ or verifying properties of polynomials used in lattice-based and code-based cryptosystems) relies on the same reduction-mod-$p$ philosophy that powers this proof. In computational algebraic geometry and symbolic computation, content/primitive-part decompositions are a standard preprocessing step (analogous to factoring out gcds before running the Euclidean algorithm) that keeps coefficient growth under control during polynomial GCD computations (e.g. in the modular GCD and subresultant algorithms).

:::quiz[If $f,g \in \mathbb{Z}[x]$ are primitive, which of the following must also be primitive?]

• $f+g$
• $2f$
• $fg$ ✓
• $f(x^2)$ alone without further hypotheses on $g$

:::quiz[What is $c(6x^2+9x+15)$ in $\mathbb{Z}[x]$, up to sign?]

• 1
• 2
• 3 ✓
• 6

:::quiz[In $\mathbb{Z}[\sqrt{-5}]$, why can the standard proof of Gauss's Lemma break down?]

• Because $\mathbb{Z}[\sqrt{-5}]$ has no field of fractions
• Because some irreducible elements are not prime, so $R/pR$ need not be a domain ✓
• Because $\mathbb{Z}[\sqrt{-5}]$ is finite
• Because polynomials over $\mathbb{Z}[\sqrt{-5}]$ cannot be multiplied

:::faq[How does Gauss's Lemma relate to $R[x]$ being a UFD?] Gauss's Lemma is the crucial multiplicative ingredient in the standard proof that $R[x]$ is a UFD whenever $R$ is. The overall strategy factors any $f \in R[x]$ as content times a primitive part, factors the primitive part in $F[x]$ (a UFD since $F$ is a field and $F[x]$ is a Euclidean domain, hence a PID, hence a UFD), and then uses Gauss's Lemma to show that the irreducible factors obtained in $F[x]$ can be rescaled to primitive polynomials in $R[x]$ whose product still matches, giving a genuine factorization in $R[x]$ with the same uniqueness properties.

Ask the tutor: Can you walk me through how Gauss's Lemma is used inside the proof that $R[x]$ is a UFD whenever $R$ is?

Think of the content of a polynomial as measuring how much of a "common scalar factor" is hiding inside its coefficients, exactly the way $\gcd(6,10,4) = 2$ measures the shared factor of $6x^2+10x+4$. Once you divide that scalar out, you are left with a primitive polynomial — one whose coefficients are, prime by prime, "spread out" enough that no single prime divides them all. The subtlety Gauss's Lemma addresses is this: when you multiply two polynomials, could a new common prime factor magically appear in the product, one dividing every coefficient of $fg$ even though it divided no single coefficient of $f$ or of $g$ across the board? The lemma says no. The mechanism is a "worst term" argument done modulo a prime $p$: if $p$ divided every coefficient of $fg$, look at the lowest-degree coefficient of $f$ not divisible by $p$, and the lowest-degree coefficient of $g$ not divisible by $p$; their product contributes a term of $fg$ that $p$ cannot kill, because all the terms that could interfere are individually divisible by $p$. A convenient way to visualize this is to picture the coefficients of $f$ and $g$ arranged on two number lines, with "poison marks" at multiples of $p$, and the corresponding convolution sum for a coefficient of $fg$. The first unmarked coefficients on each line multiply to give an unmarked contribution that no other term in the convolution can cancel modulo $p$.

This local, prime-by-prime picture is the entire content of the proof: Gauss's Lemma is really a statement about the ring $(R/pR)[x]$ being an integral domain when $p$ is prime, applied to a well-chosen reduction map.

Example 1 (asymmetric integer polynomials). Let

Compute contents: $c(f) = \gcd(4,6,10,2) = 2$, and $c(g) = \gcd(9,15,6) = 3$. So

By Gauss's Lemma, $f_1 g_1$ must be primitive. Let's verify by direct multiplication:

Expanding term by term:

Summing coefficients by degree:

• $x^5$: $6$
• $x^4$: $10+9=19$
• $x^3$: $4+15+15=34$
• $x^2$: $6+25+3=34$
• $x^1$: $10+5=15$
• $x^0$: $2$

So $f_1g_1 = 6x^5+19x^4+34x^3+34x^2+15x+2$, and $\gcd(6,19,34,34,15,2)=1$ (e.g. $19$ is odd, and prime $19$ does not divide $6$), confirming primitivity as guaranteed. Now by the Corollary, $c(fg) = c(f)c(g) = 6$ up to a unit, and indeed

whose coefficient gcd is $\gcd(36,114,204,204,90,12) = 6$, matching $c(f)c(g)=6$ exactly (in $\mathbb{Z}$, up to the unit $\pm 1$, and we may take the positive representative).

Example 2 (rational coefficients cleared via content — the standard application). Consider

Clear denominators by multiplying by $6$ (the lcm of denominators):

Now $c(4x^2+5x+3) = \gcd(4,5,3) = 1$, so $4x^2+5x+3$ is already primitive; write $h = \tfrac16 (4x^2+5x+3)$. Suppose we want to factor $h$ over $\mathbb{Q}$, and suppose we suspect $h = p(x) q(x)$ with $p,q \in \mathbb{Q}[x]$ monic-adjustable rational factors, say degree $1$ each times a rational constant. Each of $p,q$ can be written as (rational scalar)$\times$(primitive integer polynomial), say $p = \alpha p_1$, $q = \beta q_1$ with $p_1, q_1 \in \mathbb{Z}[x]$ primitive and $\alpha,\beta \in \mathbb{Q}^\times$. Then

and by Gauss's Lemma $p_1q_1$ is primitive in $\mathbb{Z}[x]$. Comparing with $h = \tfrac16(4x^2+5x+3)$ where $4x^2+5x+3$ is primitive, uniqueness of the content decomposition (up to units $\pm 1$ in $\mathbb{Z}$) forces $p_1 q_1 = \pm(4x^2+5x+3)$ and $\alpha\beta = \pm \tfrac16$. So any rational factorization of $h$ reduces entirely to an integer factorization of $4x^2+5x+3$. Indeed $4x^2+5x+3$ factors over $\mathbb{Z}$ as... testing: discriminant $25-48=-23<0$, so it is irreducible over $\mathbb{Q}$ (no real roots, and being degree $2$, reducibility over $\mathbb{Q}$ would require a rational, hence real, root). Hence $h$ itself is irreducible over $\mathbb{Q}$, and this was decided purely by an integer gcd/discriminant computation — exactly the payoff of Gauss's Lemma: rational factorization questions become integer factorization questions.