Polynomial Rings

Polynomial rings are the algebraic model behind algebraic geometry: the zero set of an ideal $I \subset k[x_1,\dots,x_n]$ defines an affine variety, and Hilbert's Nullstellensatz translates geometric statements about these zero sets into ring-theoretic statements about $I$ and its radical — all resting on $k[x_1,\dots,x_n]$ being a well-defined, well-behaved commutative ring. In cryptography, polynomial rings over finite fields, $\mathbb{F}_p[x]$, and their quotients underlie AES's S-box construction and Reed–Solomon error-correcting codes, where the domain property of $\mathbb{F}_p[x]$ guarantees unique factorization of codewords' generator polynomials. In numerical analysis and computer algebra systems (Mathematica, Sage, Macaulay2), polynomial arithmetic is implemented literally as coefficient-array convolution, exactly as in the definition given here, with degree tracked as an array index for efficient multiplication algorithms (schoolbook, Karatsuba, FFT-based). In combinatorics, formal power series and generating functions extend $R[x]$ to $R[[x]]$, and the same convolution-based multiplication rule computes convolutions of counting sequences. In control theory and signal processing, transfer functions are ratios of polynomials in a ring built exactly this way, and stability analysis depends on degree and factorization properties transferred from the coefficient field $\mathbb{R}$ or $\mathbb{C}$.

:::quiz[In $R[x]$ with $R$ an integral domain, what is $\deg(f \cdot 1)$ for a nonzero polynomial $f$ of degree $5$?]

• 0, since multiplying by the identity resets degree
• 5, since $\deg(1) = 0$ and degrees add ✓
• undefined, since $1$ is a unit
• 6, one more than $\deg f$

:::quiz[Which of the following is the correct reason $R[x]^\times = R^\times$ requires $R$ to have no zero divisors?]

• Without this hypothesis, $R[x]$ might not contain $x$ at all
• Without this hypothesis, the degree of a product can be less than the sum of degrees, allowing positive-degree units ✓
• Without this hypothesis, $R$ would not embed in $R[x]$
• Without this hypothesis, $1_R$ might equal $0_R$

:::quiz[In $\mathbb{Z}_4[x]$, which polynomial is a unit of positive degree, illustrating the counterexample?]

• $x + 1$
• $2x$
• $2x + 1$ ✓
• $3x + 2$

:::quiz[If $R$ is an integral domain, which statement about $R[x_1,x_2]$ (two variables) is justified by iterating the domain-transfer theorem?]

• $R[x_1,x_2]$ is a field
• $R[x_1,x_2]$ is an integral domain, since $R[x_1]$ is a domain and $R[x_1,x_2] = (R[x_1])[x_2]$ ✓
• $R[x_1,x_2]$ has the same units as $R[x_1]$ only, not $R$
• $R[x_1,x_2]$ cannot contain zero divisors even if $R$ does

:::faq[Why do we define $R[x]$ using coefficient sequences instead of just saying "formal expressions in $x$"?] The sequence definition removes all ambiguity about what "$x$" is and what relations it satisfies. If we simply declared $x$ a "formal symbol," we would still need to specify precisely when two such expressions are equal and how to multiply them — and the honest way to do that is to define equality as componentwise equality of coefficients and multiplication via convolution. Once this is done rigorously, the familiar notation $a_0+a_1x+\cdots+a_nx^n$ can be recovered as a theorem (as shown in the definition block), not assumed. This mirrors how $\mathbb{C}$ is rigorously built as pairs $(a,b)\in\mathbb{R}^2$ before we write $a+bi$.

:::faq[Does $R[x]$ ever become a field, even when $R$ is a field?] No. Even if $R = k$ is a field, $k[x]$ is never a field, because $x$ has no multiplicative inverse: if $xg = 1$ for some $g \in k[x]$, then by the degree formula (valid since $k$ has no zero divisors) $\deg(xg) = 1 + \deg g \ge 1$, but $\deg(1) = 0$, a contradiction. So $x$ is a nonzero non-unit, and $k[x]$ is an integral domain but not a field. This is exactly analogous to $\mathbb{Z}$ being a domain but not a field, and it motivates forming the Field of Fractions of $k[x]$, namely the field of rational functions $k(x)$.

:::faq[Why does the integral domain property of $R[x]$ matter for Gauss's Lemma?] Gauss's Lemma compares factorization of a polynomial in $\mathbb{Z}[x]$ to factorization in $\mathbb{Q}[x]$, and a first step in that theory is knowing $\mathbb{Z}[x]$ is an integral domain (so factorization questions are meaningful — no zero divisors muddying products) and that its units are only $\pm 1$ (so "irreducible" is a well-posed notion up to unit multiples). Both facts are immediate corollaries of the theorem proved here, applied with $R = \mathbb{Z}$, and they are exactly the starting hypotheses Gauss's Lemma builds upon.

Ask the tutor: How does the degree formula proved here get used to show $k[x]$ is a Euclidean domain for a field $k$?

:::definition[The polynomial ring $R[x]$] Let $R$ be a commutative ring with identity $1_R$. A polynomial over $R$ is a sequence

such that $a_i = 0$ for all but finitely many $i$ (finite support). The set of all such sequences is denoted $R[x]$. Define addition componentwise:

Define multiplication by the convolution rule:

Both operations preserve finite support (a routine check), so $R[x]$ is closed under $+$ and $\cdot$. One verifies directly, using only the ring axioms of $R$, that $(R[x], +, \cdot)$ is a commutative ring with additive identity $(0,0,\dots)$ and multiplicative identity $(1_R, 0, 0, \dots)$. Let $x := (0, 1, 0, 0, \dots) \in R[x]$. A direct computation from the convolution rule shows

for $n \ge 1$ (the $1$ in position $n$), and $x^0 = 1_R[x]$ by convention. Identify $R$ with the subring of constant polynomials

via the injective ring homomorphism $a \mapsto (a, 0, 0, \dots)$. Under this identification, every polynomial $f = (a_0,\dots,a_n,0,0,\dots)$ with $a_n \neq 0$ can be written, using the ring operations of $R[x]$ alone, as

This is the familiar notation for polynomials, now justified as a derived fact rather than an assumption. If $f \neq 0$ and $f = a_0 + \cdots + a_n x^n$ with $a_n \neq 0$, we call $n = \deg f$ the degree of $f$, and $a_n$ its leading coefficient; $f$ is monic if $a_n = 1_R$. The zero polynomial is assigned no degree (some authors use $\deg 0 = -\infty$, a convention we mention but do not rely on).

The construction above looks bureaucratic, but the point is precisely to earn the notation "$a_0 + a_1x + \cdots + a_nx^n$" rather than assume it. Coefficient sequences are unambiguous: two polynomials are equal exactly when all their coefficients agree, term by term. This resolves any worry about $x$ being some mysterious "variable" — it is simply a specific, fixed element of $R[x]$, namely $(0,1,0,0,\dots)$, and all algebraic manipulation with $x$ is just ring arithmetic in $R[x]$. Picture $R[x]$ as an infinite-dimensional free $R$-module with basis $1, x, x^2, x^3, \dots$, where multiplication is dictated by $x^i \cdot x^j = x^{i+j}$ and $R$-linearity. The degree of a nonzero polynomial records the highest basis vector with nonzero coefficient — the "outermost" surviving term. Multiplying two polynomials convolves their coefficient sequences, and the only way the top-degree term of the product can vanish is if the two leading coefficients multiply to zero in $R$. This is exactly the scenario ruled out when $R$ has no zero divizors, and it is the single mechanism behind the degree-addition formula, the units theorem, and the domain-transfer theorem proved below.

:::theorem[Degree, units, and domain transfer for $R[x]$] Let $R$ be a commutative ring with identity, and suppose $R$ has no zero divisors (equivalently, since $R$ is commutative with $1 \ne 0$, $R$ is an integral domain). Let $f, g \in R[x]$ be nonzero polynomials. Then: 1. (Degree formula) $fg \neq 0$ and

2. (Units) $u \in R[x]$ is a unit if and only if $u$ is a (constant) unit of $R$, i.e.

3. (Domain transfer) $R[x]$ is an integral domain.

Example 1: $\mathbb{Z}[x]$, an asymmetric product. Let $R = \mathbb{Z}$, an integral domain. Take

so $\deg f = 4$ with leading coefficient $3$, and $\deg g = 3$ with leading coefficient $-2$. Multiply directly:

Expand term by term:

Collecting like terms:

The leading term is $-6x^7$, consistent with the theorem's prediction: leading coefficient $3\cdot(-2) = -6 \neq 0$ since $\mathbb{Z}$ has no zero divisors, and

Also, $fg \ne 0$, confirming $\mathbb{Z}[x]$ has no zero divisors here, and neither $f$ nor $g$ (both of positive degree) could ever be a unit: indeed $\mathbb{Z}[x]^\times = \mathbb{Z}^\times = \{1,-1\}$ by part 2 of the theorem.

Example 2: failure of degree-additivity without the domain hypothesis, inside $\mathbb{Z}_{12}[x]$. Let $R = \mathbb{Z}_{12}$, which is not an integral domain since $3 \cdot 4 = 12 = 0$ with $3, 4 \neq 0$. Take

Here $\deg f = 2$ (leading coeff $3$), $\deg g = 3$ (leading coeff $4$). Compute the product's degree-$5$ coefficient using the convolution formula: it is the product of leading coefficients,

So the $x^5$-term of $fg$ vanishes even though both $f,g$ have positive degree contributing to it — exactly the failure mechanism the theorem's proof isolates in Step 2. Let us find the true degree of $fg$ by full expansion. Write $f = 3x^2+x+1$, $g=4x^3+2x+5$:

Reduce mod 12 and collect:

So

Here $\deg(fg) = 4 \neq 2+3=5$: the degree formula fails, exactly because $\mathbb{Z}_{12}$ has zero divisors. This example is placed here precisely to show the theorem's hypothesis (no zero divisors in $R$) is not decorative — it is used essentially in Step 2 of the proof.