Gaussian quadrature

Gaussian quadrature
Paolostar · CC BY-SA 4.0 · source
Name	Gaussian quadrature
Field	Numerical analysis
Introduced	19th century
Contributors	Carl Friedrich Gauss, Adrien-Marie Legendre, François Vieillefosse, Carl Jacobi, Carl Gustav Jacob Jacobi

Gaussian quadrature is a family of numerical integration techniques that select optimal evaluation points and weights to exactly integrate polynomials up to a maximal degree. Developed in the 19th century and associated with figures such as Carl Friedrich Gauss and Adrien-Marie Legendre, it underpins many modern computational methods in applied mathematics and scientific computing. Gaussian quadrature methods exploit properties of orthogonal polynomials and have influenced numerical techniques used in physics, engineering, and finance.

Introduction

Gaussian quadrature provides an n-point rule to approximate definite integrals by a weighted sum of function values at carefully chosen nodes. The method achieves exactness for polynomials of degree up to 2n−1 by solving moment conditions tied to orthogonality relations discovered by researchers like Karl Weierstrass and Émile Borel. It contrasts with Newton–Cotes rules attributed to Isaac Newton and Roger Cotes and finds practical use in contexts ranging from spectral methods used in John von Neumann-influenced numerical linear algebra to finite element integrations relevant to Sophus Lie-inspired symmetry analyses.

Theory and Derivation

Derivation begins with the projection of functions onto a polynomial basis that is orthogonal with respect to a weight function on an interval. Results rely on classical theorems from analysis such as those by Bernhard Riemann and Pierre-Simon Laplace on quadrature and approximation, and on determinant identities related to Carl Gustav Jacobi. The algebraic formulation leads to a system of equations for nodes and weights tied to moments; solving these uses techniques from linear algebra popularized by David Hilbert and John von Neumann, and determinants and eigenvalue problems studied by Arthur Cayley and James Joseph Sylvester.

Orthogonal Polynomials and Nodes

Nodes of Gaussian rules are the roots of orthogonal polynomials associated with a weight function on an interval. Classical families include polynomials named after Adrien-Marie Legendre, Pafnuty Chebyshev, Carl Gustav Jacobi, and Heinrich Eduard Heine, each yielding specific node distributions. The connection between roots and three-term recurrence relations was clarified by analysts like Thomas Stieltjes and further formalized in spectral theory by John von Neumann and Marshall Stone. Computation of nodes via eigenvalue problems leverages matrices studied by Issai Schur and Erhard Schmidt.

Weight Functions and Types (Gauss–Legendre, Gauss–Chebyshev, etc.)

Different weight functions give rise to distinct Gaussian rules: the rule on [−1,1] with weight 1 is associated with Adrien-Marie Legendre and yields the Gauss–Legendre rule; weights (1−x^2)^{-1/2} and (1−x^2)^{1/2} lead to Chebyshev-related rules tied to Pafnuty Chebyshev. Jacobi polynomials parameterized by α and β produce Gauss–Jacobi rules linked to Carl Gustav Jacobi, while Laguerre and Hermite families connected to Edmond Laguerre and Charles Hermite produce rules on semi-infinite and infinite intervals respectively. These classical families are central to applications in contexts influenced by Joseph Fourier methods, quantum mechanics formulations from Erwin Schrödinger, and orthogonality theory advanced by Gustav Doetsch.

Construction and Algorithms

Practical construction of nodes and weights uses recurrence relations and orthogonalization algorithms developed from work by Carl Friedrich Gauss and modern numerical linear algebra methods by Gene H. Golub and William Kahan. Algorithms include Golub–Welsch methods converting recurrence coefficients to symmetric tridiagonal matrices whose eigenvalues are nodes, and iterative root-finding techniques inspired by Isaac Newton and refined by John Todd and John H. Wilkinson. Stable computation employs techniques related to continued fractions studied by Leonhard Euler and matrix perturbation results by T. J. Osborne and Nicholas Higham.

Error Analysis and Convergence

Error bounds derive from remainder terms expressed via derivatives evaluated at unknown points and involve orthogonal polynomial extremal properties analyzed by G. H. Hardy and J. E. Littlewood. Convergence properties depend on analytic continuation and regularity results associated with Bernhard Riemann and approximation theorems from Andrey Kolmogorov. Gaussian rules often exhibit exponential convergence for analytic integrands, a phenomenon exploited in spectral methods linked to John von Neumann and approximation theory advanced by Sergei Natanovich Bernstein.

Applications and Extensions

Gaussian quadrature appears across numerical solutions of differential equations in methods pursued by Leonhard Euler and Simeon Denis Poisson, in computational electromagnetics influenced by James Clerk Maxwell, and in computational finance models building on ideas from Louis Bachelier and Fischer Black. Extensions include Gauss–Kronrod rules for adaptive integration introduced in computational practice influenced by H. J. Kronrod, sparse grid and Smolyak constructions tied to Sergey Smolyak, and multivariate cubature related to work by Harold Hotelling and I. M. Gelfand. Modern software implementations draw upon algorithms from numerical libraries developed in projects associated with Donald Knuth-style literate programming and HPC frameworks influenced by John Backus.

Category:Numerical analysis