Polynomials are the well-behaved family of mathematics: a finite sum of integer powers of a variable, multiplied by coefficients. p(x) = aₙxⁿ + aₙ₋₁xⁿ⁻¹ + … + a₁x + a₀. They can be added, multiplied, divided, differentiated, integrated, factored, and evaluated with no analytic difficulty whatever. The catch is that not every function is a polynomial; the gift is that every continuous function on a closed interval can be uniformly approximated by one (Weierstrass, 1885). For two thousand years, polynomials have been the workhorse of approximation theory — the language in which mathematics expresses functions when it wants to actually compute with them.
A polynomial of degree n is an expression p(x) = aₙxⁿ + aₙ₋₁xⁿ⁻¹ + … + a₀ with aₙ ≠ 0. The coefficients aᵢ are usually drawn from a ring — integers, rationals, reals, complexes, or finite fields. Operations on polynomials are elementary: addition acts on coefficients pairwise; multiplication is convolution of coefficient sequences; composition nests one polynomial inside another. Division with remainder works just as for integers, dividing one polynomial by another of lower or equal degree. The Fundamental Theorem of Algebra — proved by Gauss in his 1799 doctoral dissertation, the first of four proofs he would give over his life — states that every polynomial of degree n with complex coefficients has exactly n complex roots (counted with multiplicity). Real polynomials need not have any real roots (x² + 1 has none), but they do have all their roots somewhere in ℂ. Solvability by radicals is a finer question: degree ≤ 4 has explicit formulas (the quadratic formula; Cardano's formula for cubics; Ferrari's for quartics); degree ≥ 5 generically does not, by Galois's theorem. Polynomial interpolation (Lagrange, Newton): given n + 1 data points, there is a unique polynomial of degree ≤ n passing through all of them. Bernstein polynomials give Bezier curves, the bedrock of vector graphics. Splines — piecewise polynomials with controlled continuity at the joints — are how computer graphics, font rendering, and animation actually represent smooth shapes. The Stone-Weierstrass theorem (1937) generalizes Weierstrass approximation: in a wide class of function spaces, polynomials are dense — anything in the space can be approximated arbitrarily well by polynomials.
Computer graphics (PostScript, PDF, every vector-graphics format) represents curves as Bézier polynomials; fonts are stored as polynomial outlines. Computer algebra systems (Mathematica, SymPy) treat polynomials as primary objects, with algorithms for factorization, GCDs (Buchberger's algorithm for Gröbner bases generalizes Euclid's algorithm to polynomials in many variables), and root-finding. Reed-Solomon codes — used in QR codes, CDs, DVDs, satellite communication, and storage systems — are based on the evaluation of polynomials at multiple points. Optimization: polynomial constraints define semialgebraic sets, and entire branches of optimization (semidefinite programming, polynomial optimization) operate over them. Lattice-based post-quantum cryptography — currently being standardized by NIST as a quantum-safe replacement for RSA — relies heavily on polynomial arithmetic in finite fields. Polynomials are the bridge between algebra and analysis, the natural functions that algebra knows how to handle exactly and analysis knows how to approximate everything else with.