Kolmogorov equations

Kolmogorov equations
Name	Kolmogorov equations
Field	Probability theory; Partial differential equations
Introduced	1931
Related	Stochastic processes; Markov processes; Diffusion processes

Kolmogorov equations are fundamental partial differential equations connecting stochastic processes and analytic theory, introduced by Andrey Kolmogorov in the early 20th century. They formalize time evolution for transition probabilities and distributions of Markov processes and link to parabolic equations studied in the context of diffusion, Brownian motion, and statistical mechanics. The equations underpin mathematical treatments in probability theory, statistical physics, mathematical finance, and engineering.

Introduction

The Kolmogorov equations arise in the theory of continuous-time Markov processes, branching processes, and diffusion limits such as Brownian motion, Wiener process models, and reaction–diffusion systems, and were developed contemporaneously with work by Paul Lévy, Norbert Wiener, Émile Borel, Sergio Fubini, and Richard Courant. They connect to classical results from Andrey Kolmogorov's contemporaries including Aleksandr Lyapunov, Andrei Markov, André Weil, Élie Cartan, and later advances by Kiyoshi Itō, Joseph Doob, William Feller, and Henry McKean. The forward and backward formulations complement techniques used in the study of generators associated with semigroups by Einar Hille, Rudolf Phillips, Klaus Friedrichs, and the spectral theory of John von Neumann.

Forward and Backward Kolmogorov Equations

The forward equation, often called the Fokker–Planck equation in physics, describes evolution of probability densities for diffusions and relates to the infinitesimal generator studied by Mark Kac, Salomon Bochner, Tosio Kato, and Lars Hörmander. The backward equation provides expectations of functionals and connects to martingale problems formulated by Donald Burkholder, Burton Kailath, H. P. McKean, and Stuart S. Antman. In discrete settings the master equation or chemical master equation used in Gillespie algorithm contexts links to work by Rudolf Peierls, Sin-Itiro Tomonaga, and George Uhlenbeck. Both equations interact with semigroup theory developed by Einar Hille and Rudolf Phillips and operator methods advanced by Marshall Stone and Nelson Dunford.

Derivation and Mathematical Properties

Derivations employ limit theorems such as the central limit theorem associated with Andrey Kolmogorov and Paul Lévy, coupling arguments used by Kiyoshi Itō, and martingale representations from Joseph Doob and Meyer Meyer (Jacques Meyer). Mathematical structure involves generators which are second-order elliptic operators treated in the frameworks of Lax–Milgram theorem, Sobolev spaces associated with Laurent Schwartz and Sergei Sobolev, and hypoellipticity criteria of Lars Hörmander. Existence and uniqueness results appeal to maximum principles related to Eberhard Hopf, analytic semigroups developed by Amnon Pazy, and elliptic regularity theorems advanced by Ennio de Giorgi, John Nash, and Sergiu Klainerman. Boundary conditions reference classical problems studied by Siméon Denis Poisson, Augustin-Louis Cauchy, and Joseph Fourier.

Applications in Stochastic Processes and PDEs

Kolmogorov equations apply to models in mathematical finance pioneered by Fischer Black, Myron Scholes, and Robert Merton for option pricing, to population dynamics studied by Ronald Fisher, J. B. S. Haldane, and Sewall Wright, and to chemical kinetics modeled by Gilbert N. Lewis and computational methods by Dan Gillespie. They underpin diffusion approximations in queueing theory from John Kingman and in epidemiology linked to work by Daniel Bernoulli and Andrei Kolmogorov's collaborators. Connections to statistical mechanics relate to the foundational studies of Ludwig Boltzmann, James Clerk Maxwell, and Josiah Willard Gibbs, while applications in control theory tie to research by Richard Bellman, Rudolf Kalman, and Lotfi Zadeh.

Solutions and Numerical Methods

Analytic solutions are available in special geometries studied by S. R. Srinivasa Varadhan, Peter Lax, and Elliott H. Lieb using transform techniques of Joseph Fourier and spectral methods of David Hilbert and John von Neumann. Numerical approaches include finite difference schemes developed in the tradition of Richard Courant and K. O. Friedrichs, finite element methods associated with J. H. Argyris and Ivo Babuška, stochastic simulation algorithms influenced by Daniel Gillespie and Monte Carlo methods of Stanislaw Ulam and Nicholas Metropolis, and particle filter techniques related to Gene Golub and Andrew Gelman. Stability, consistency, and convergence analyses draw on criteria by Courant–Friedrichs–Lewy and work on error estimates by G. D. Birkhoff and Andrey Kolmogorov's successors.

Examples and Important Special Cases

Classical examples include the heat equation for Brownian motion traced by Joseph Fourier and Lord Kelvin, the Ornstein–Uhlenbeck process introduced by Leonard Ornstein and George E. Uhlenbeck, birth–death processes analyzed by W. Feller and Samuel Karlin, the chemical master equation in reaction networks used by Daniel Gillespie and Nicolas G. Van Kampen, and the Black–Scholes PDE in finance by Fischer Black and Myron Scholes. Other special cases appear in branching processes studied by Galton–Watson and spatial ecology models linked to A. J. Lotka and Vito Volterra, and in homogenization results by V. V. Zhikov and Grigory Barenblatt.

Category:Probability theory