Polynomials are among the few functions that can be understood from several viewpoints at once. Algebraically, a polynomial is a finite linear combination of powers of a variable. Analytically, it is a smooth function whose local behaviour is controlled by finitely many coefficients. Geometrically, the graph of a real polynomial is a curve with a rigid large-scale shape. This mixture of algebraic finiteness and analytic flexibility is why polynomials appear throughout [Function](/page/Function), [Derivative](/page/Derivative), [Limit](/page/Limit), [Continuity](/page/Continuity), [approximation theory](/page/Approximation%20Theory), and differential equations. The finiteness condition is the central point. Infinite series such as [power series](/page/Power%20Series) can model a much larger class of functions, but they require convergence questions before they define functions. A polynomial does not: after choosing a coefficient ring or field and a variable, only finitely many arithmetic operations are involved. That makes polynomials the first robust class of functions in which algebraic operations, roots, derivatives, integrals, and asymptotic behaviour can be studied together. ## Definition Before treating polynomials as functions, it is useful to separate the symbolic object from the function it may define. The expression $x^2-1$ is not merely a rule on [real numbers](/page/Real%20Numbers); it can be manipulated over $\mathbb{Q}$, $\mathbb{R}$, $\mathbb{C}$, or a more general [Ring](/page/Ring). This distinction matters because different coefficient systems can change factorisation, roots, and even whether two polynomial expressions define the same function. The formal definition records exactly the finite coefficient data that survives all these changes of setting. [definition: Polynomial] Let $R$ be a commutative ring with identity. A polynomial in one variable $x$ with coefficients in $R$ is a formal finite sum \begin{align*} p = \sum_{i=0}^{n} a_i x^i, \end{align*} where $n \in \mathbb{N} \cup \{0\}$, each $a_i \in R$, and all but finitely many coefficients are zero. The set of all such polynomials is denoted $R[x]$. [/definition] [example: A Quadratic as a Formal Polynomial] In $\mathbb{R}[x]$, the expression \begin{align*} p=x^2-3x+2 \end{align*} is the finite coefficient list $a_0=2$, $a_1=-3$, $a_2=1$, and $a_i=0$ for all $i\ge 3$. Its degree is $2$, and its leading coefficient is $1$, so it is monic. When it is evaluated as a real function, the input $t=3$ gives \begin{align*} \tilde p(3)=3^2-3\cdot 3+2=2. \end{align*} [/example] The phrase "formal finite sum" means that the coefficients are part of the data. Addition and multiplication are defined by collecting coefficients, rather than by evaluating at a point. The notation $p(x)$ is often used informally for such an element of $R[x]$, but evaluation at a specific input is a separate operation introduced below. Multivariable polynomials are the standard extension obtained by allowing finitely many monomials in variables $x_1, \dots, x_m$; this page focuses on the one-variable case because degree, roots, and polynomial functions already show the essential issues there. To compare polynomials by complexity, to run induction on them, or to state how roots constrain them, we need to know the largest power whose coefficient is actually present. [definition: Degree of a Polynomial] Let $R$ be a commutative ring with identity, and let \begin{align*} p(x) = \sum_{i=0}^{n} a_i x^i \in R[x]. \end{align*} If $p \ne 0$, the degree of $p$ is \begin{align*} \deg p = \max\{i \in \{0, \dots, n\} : a_i \ne 0\}. \end{align*} [/definition] To predict division, factorisation, and growth at infinity, a polynomial needs a distinguished top coefficient as well as a highest power. That coefficient is the part of the polynomial that survives after all lower powers become negligible or after a [division algorithm](/theorems/725) normalises the top term. [definition: Leading Coefficient] Let $R$ be a commutative ring with identity, and let $p \in R[x]$ be nonzero. If $\deg p = d$ and \begin{align*} p(x) = \sum_{i=0}^{d} a_i x^i \end{align*} with $a_d \ne 0$, then $a_d$ is the leading coefficient of $p$. [/definition] To compare polynomials up to multiplication by an invertible scalar, it is useful to choose a standard representative whenever possible. The simplest normalisation is to require the leading coefficient to be the multiplicative identity. [definition: Monic Polynomial] Let $R$ be a commutative ring with identity. A nonzero polynomial $p \in R[x]$ is monic if its leading coefficient is $1_R$. [/definition] In analysis, a polynomial is often used as a function on $\mathbb{R}$ or $\mathbb{C}$. That requires evaluation: a formal expression becomes a rule that sends an input to the corresponding finite arithmetic combination of powers. The next definition names this functional interpretation while keeping it separate from the formal polynomial. [definition: Polynomial Function] Let $k$ be a field, and let $p \in k[x]$ be given by \begin{align*} p(x) = \sum_{i=0}^{n} a_i x^i. \end{align*} The polynomial function associated to $p$ is the map $\tilde{p}: k \to k$ defined by \begin{align*} \tilde{p}(t)=\sum_{i=0}^{n} a_i t^i. \end{align*} [/definition] Over infinite fields such as $\mathbb{R}$ and $\mathbb{C}$, distinct formal polynomials define distinct polynomial functions. Over finite fields this may fail, so the symbolic and functional viewpoints should not be silently identified in every setting. ## Equivalent Characterisations For real or complex analysis, the defining finite sum is often the most convenient form. Another way to recognise polynomials is through their derivatives: after enough differentiations, a polynomial becomes zero. This reformulation connects polynomials to [Taylor's theorem](/theorems/827) and to local approximation by finite jets. [quotetheorem:8293] This result explains why polynomials are the model finite-order smooth functions. A general smooth function can have derivatives of every order that carry new information, while a polynomial stores all of its information in finitely many derivatives at a single point. To use this as a test for whether a smooth function is polynomial, we need a converse statement on domains where local information cannot split into unrelated pieces. [quotetheorem:8294] The interval hypothesis prevents the function from being independently polynomial on disconnected pieces. This is a useful warning: derivative conditions are local, while a single polynomial formula is global on the domain being considered. For many analytic and numerical applications, one does not need all polynomials at once; instead, one fixes a maximum degree and obtains a finite-dimensional space. [definition: Polynomial Space of Bounded Degree] Let $k$ be a field and let $d \in \mathbb{N} \cup \{0\}$. The space of polynomials over $k$ of degree at most $d$ is \begin{align*} \mathcal{P}_d(k)=\{p \in k[x] : p=0 \text{ or } \deg p \le d\}. \end{align*} [/definition] To use $\mathcal{P}_d(k)$ computationally, we need a coordinate system for its elements. Without a basis, a bounded-degree polynomial is still a formal expression whose coefficients are hard to compare or count. The monomials $1,x,x^2,\dots,x^d$ remove this ambiguity: each polynomial of degree at most $d$ should have exactly one list of $d+1$ coefficients, so the space should behave like a $d+1$-dimensional [vector space](/page/Vector%20Space). [quotetheorem:8295] This theorem is the reason a polynomial of degree at most $d$ is determined by $d+1$ independent scalar conditions, provided those conditions are chosen without degeneracy. Interpolation is the most familiar instance. ## Standard Examples The first example shows the basic analytic behaviour of a real polynomial: derivatives eventually vanish, but the lower derivatives still encode meaningful information. [example: Cubic Polynomial and Its Derivatives] Consider the real polynomial function $p: \mathbb{R} \to \mathbb{R}$ defined by \begin{align*} p(x)=2x^3-3x+5. \end{align*} The largest power of $x$ with nonzero coefficient is $x^3$, so $\deg p=3$, and the coefficient of that highest power is $2$, so the leading coefficient is $2$. Using linearity of differentiation and the power rule, \begin{align*} p'(x)=\frac{d}{dx}(2x^3)-\frac{d}{dx}(3x)+\frac{d}{dx}(5)=2\cdot 3x^2-3\cdot 1+0=6x^2-3. \end{align*} Differentiating again gives \begin{align*} p''(x)=\frac{d}{dx}(6x^2)-\frac{d}{dx}(3)=6\cdot 2x+0=12x. \end{align*} The third derivative is \begin{align*} p^{(3)}(x)=\frac{d}{dx}(12x)=12, \end{align*} and the fourth derivative is \begin{align*} p^{(4)}(x)=\frac{d}{dx}(12)=0. \end{align*} Since the derivative of the zero function is again the zero function, every derivative of order at least $4$ is zero on $\mathbb{R}$. Thus the cubic has nonzero derivative information through order $3$, but no derivative information remains after that. [/example] The next example marks a boundary between formal polynomials and polynomial functions. It is invisible over $\mathbb{R}$, but it appears as soon as the coefficient field is finite. [example: Distinct Formal Polynomials with the Same Function] Let $k=\mathbb{F}_2=\{0,1\}$, where addition and multiplication are taken modulo $2$. In $k[x]$, consider \begin{align*} p(x)=x^2+x, \qquad q(x)=0. \end{align*} These are distinct formal polynomials because the coefficient of $x^2$ in $p$ is $1$, while the coefficient of $x^2$ in $q$ is $0$. Now compare the associated functions $k \to k$. At the input $0$, \begin{align*} p(0)=0^2+0=0+0=0. \end{align*} Also, \begin{align*} q(0)=0. \end{align*} At the input $1$, \begin{align*} p(1)=1^2+1=1+1=0, \end{align*} because $1+1=0$ in $\mathbb{F}_2$. Also, \begin{align*} q(1)=0. \end{align*} Since the only elements of $\mathbb{F}_2$ are $0$ and $1$, the two associated functions agree at every input: \begin{align*} \tilde{p}(t)=\tilde{q}(t) \quad \text{for every } t \in \mathbb{F}_2. \end{align*} Thus $p$ and $q$ define the same polynomial function on $\mathbb{F}_2$ even though $p \ne q$ as elements of $\mathbb{F}_2[x]$, so the formal polynomial and the function it induces cannot be identified in every coefficient field. [/example] A different boundary appears when a function resembles a polynomial near a point but is not globally a polynomial. Taylor polynomials are approximating objects, not the same thing as the function being approximated. [example: Taylor Polynomial Is Not the Original Function] Let $f: \mathbb{R} \to \mathbb{R}$ be $f(x)=e^x$. Since $\frac{d}{dx}e^x=e^x$, induction gives $f^{(m)}(x)=e^x$ for every $m \in \mathbb{N}$, so \begin{align*} f(0)=f'(0)=f''(0)=f^{(3)}(0)=1. \end{align*} The degree-$3$ Taylor polynomial at $0$ is therefore \begin{align*} T_3(x)=f(0)+f'(0)x+\frac{f''(0)}{2!}x^2+\frac{f^{(3)}(0)}{3!}x^3=1+x+\frac{x^2}{2}+\frac{x^3}{6}. \end{align*} This is a polynomial because it is a finite linear combination of $1,x,x^2,x^3$. The function $f$ is not a polynomial. Indeed, for every integer $m \ge 0$, \begin{align*} f^{(m)}(x)=e^x>0 \end{align*} for every $x \in \mathbb{R}$, so no derivative order vanishes identically. By *Polynomial Derivatives Vanish After the Degree*, every polynomial has all sufficiently high derivatives equal to zero, hence $e^x$ cannot be a polynomial function on $\mathbb{R}$. By *Taylor's Theorem with Remainder*, for each $x$ there is a point $\xi$ between $0$ and $x$ such that \begin{align*} e^x-T_3(x)=\frac{f^{(4)}(\xi)}{4!}x^4=\frac{e^\xi}{24}x^4. \end{align*} If $|x|\le 1$, then $\xi$ lies between $-1$ and $1$, so $e^\xi \le e$ and \begin{align*} |e^x-T_3(x)|=\frac{e^\xi}{24}|x|^4 \le \frac{e}{24}|x|^4. \end{align*} Thus $e^x-T_3(x)=O(x^4)$ as $x \to 0$. Finally, $T_3^{(4)}(x)=0$ for every $x$, while $f^{(4)}(x)=e^x>0$ for every $x$. If $f$ and $T_3$ were equal on an open interval, differentiating the equality four times on that interval would give $e^x=0$ there, which is impossible. Thus the Taylor polynomial approximates $e^x$ to fourth order at $0$, but it is not the same function on any open interval. [/example] This example is the analytic reason polynomials are not merely elementary functions. They are the finite approximants against which more complicated functions are measured. ## Algebraic and Analytic Structure ### Closure and Degree The algebraic power of polynomials comes from closure: adding or multiplying polynomials gives another polynomial. This permits polynomial rings to support algebraic operations while remaining inside the same class of finite expressions. [quotetheorem:8296] The multiplication statement depends on finiteness: each coefficient of a product is a finite sum of products of coefficients. Beyond closure, degree should behave like a size measurement under multiplication, but that requires ruling out cancellation of the leading term by zero divisors. [quotetheorem:3234] The integral-domain hypothesis is necessary. If leading coefficients can multiply to zero, the expected leading term can disappear. [example: Degree Formula Fails with Zero Divisors] Work in $R=\mathbb{Z}/4\mathbb{Z}$, so two integers represent the same coefficient when they have the same remainder modulo $4$. Let \begin{align*} p(x)=2x+1, \qquad q(x)=2x+1. \end{align*} The coefficient of $x$ in each polynomial is $2$, and $2 \ne 0$ in $\mathbb{Z}/4\mathbb{Z}$, so $\deg p=1$ and $\deg q=1$. Multiplying in $R[x]$ gives \begin{align*} p(x)q(x)=(2x+1)(2x+1). \end{align*} Expanding by distributivity, \begin{align*} (2x+1)(2x+1)=(2x)(2x)+(2x)(1)+1(2x)+1\cdot 1. \end{align*} Each term is \begin{align*} (2x)(2x)=4x^2, \qquad (2x)(1)=2x, \qquad 1(2x)=2x, \qquad 1\cdot 1=1. \end{align*} Therefore \begin{align*} p(x)q(x)=4x^2+2x+2x+1=4x^2+4x+1. \end{align*} In $\mathbb{Z}/4\mathbb{Z}$, the coefficient $4$ is equal to $0$, so \begin{align*} 4x^2+4x+1=0x^2+0x+1=1. \end{align*} Thus $pq$ is the constant polynomial $1$, so $\deg(pq)=0$, while \begin{align*} \deg p+\deg q=1+1=2. \end{align*} The expected top-degree term disappears because the leading coefficient product is $2\cdot 2=4=0$ in $\mathbb{Z}/4\mathbb{Z}$, showing why the degree formula for products requires the absence of zero divisors. [/example] ### Roots, Factors, and Uniqueness Roots are where the symbolic and functional viewpoints meet. Evaluating a polynomial at a scalar asks whether the formal expression vanishes at that scalar; factoring asks whether this vanishing is explained by a linear factor. We first name the value-based notion, then relate it to divisibility. [definition: Root of a Polynomial] Let $R$ be a commutative ring with identity, let $p \in R[x]$, and let $a \in R$. The element $a$ is a root of $p$ if the evaluation of $p$ at $a$ is zero: \begin{align*} p(a)=0_R. \end{align*} [/definition] For fields, roots are controlled by linear factors. This principle matters because it turns a question about values of a function into a question about divisibility inside the [polynomial ring](/page/Polynomial%20Ring). [quotetheorem:3235] To turn factorisation into a counting principle, each distinct root should consume at least one degree. The obstruction is that a polynomial may have repeated roots, so multiplicity can complicate factorisation, but distinct roots still force distinct linear factors over a field. This is what lets degree bound the number of different places where a nonzero polynomial can vanish. [quotetheorem:1708] This result is often the hidden engine behind uniqueness statements. To identify a bounded-degree polynomial, it is enough to test it at sufficiently many distinct points. [quotetheorem:8297] This uniqueness result is the formal reason interpolation data can specify a bounded-degree polynomial rather than merely suggest one. For example, two quadratic polynomials over $\mathbb{R}$ that agree at three distinct inputs must be the same polynomial, because their difference would be a quadratic with too many roots. The degree bound and distinctness of the sample points are both essential: without a degree bound, many different high-degree polynomials can pass through the same finite data set, and repeated sample points do not create new root constraints. This theorem is used whenever polynomial identities are checked by evaluating at enough points, and it is the bridge from the root-counting results above to interpolation and finite-dimensional approximation below. ### Calculus and End Behaviour In analysis over $\mathbb{R}$ or $\mathbb{C}$, polynomials are continuous and smooth. Because the power rule keeps derivatives inside the same class and integration raises powers by one, calculus with polynomials can be developed without leaving polynomial functions. [quotetheorem:8298] This theorem makes polynomial calculus unusually closed: differentiating lowers degree, while the displayed antiderivative raises degree by one and stays polynomial. The real-field hypothesis is doing real work in the antiderivative formula because the coefficients $\frac{a_i}{i+1}$ require division by the positive integer $i+1$ inside the coefficient field. Over settings where such division is unavailable, the same termwise formula may fail or need modification. In real analysis, however, the result lets derivative tests, Taylor polynomials, and elementary integration treat polynomials as a stable class rather than as examples that leave the category after one operation. To estimate a real polynomial far from the origin, lower powers must be compared with the highest power. The difficulty is that several terms may have different signs and sizes near ordinary points, so the polynomial's local behaviour need not reveal its eventual behaviour. Far enough from zero, however, every lower power is small relative to the leading power, and this turns the leading term into the decisive term. [quotetheorem:8299] This asymptotic statement is often the fastest way to predict whether a real polynomial tends to $\infty$, tends to $-\infty$, or has opposite behaviour at the two ends of the real line. ## Beyond and Connected Topics Polynomials are a bridge between algebra and analysis. As elements of $R[x]$, they belong to the theory of [Ring](/page/Ring), ideals, factorisation, and quotient constructions. As functions $\mathbb{R} \to \mathbb{R}$ or $\mathbb{C} \to \mathbb{C}$, they belong to [Continuity](/page/Continuity), [Derivative](/page/Derivative), integration, approximation, and complex analysis. The comparison with power series is especially important. A polynomial is a finite sum; a power series is an infinite sum whose meaning as a function depends on convergence. Every polynomial is a power series with all sufficiently high coefficients equal to zero, but many power series define functions that are not polynomials. Polynomials also organise approximation theory. Taylor polynomials approximate differentiable functions near a point, interpolation polynomials fit prescribed data, and orthogonal polynomial families provide finite-dimensional approximants inside Hilbert spaces. In [numerical analysis](/page/Numerical%20Analysis), the spaces $\mathcal{P}_d(\mathbb{R})$ are used for quadrature, finite elements, and spectral methods. In algebra, roots and factors of polynomials lead to field extensions, minimal polynomials, splitting fields, and Galois groups. A single polynomial can therefore encode both a function and an algebraic object whose roots generate new fields. In differential equations, polynomial coefficients and polynomial solutions form tractable model problems. Even when the true solution is not a polynomial, polynomial approximations are often the finite objects that can be computed, estimated, and compared. ## References [Function](/page/Function). [Derivative](/page/Derivative). [Continuity](/page/Continuity). Dummit and Foote, *Abstract Algebra* (2004). Rudin, *Principles of Mathematical Analysis* (1976). Axler, *Linear Algebra Done Right* (2015).