Partial differential equation

A partial differential equation relates an unknown function of several variables to its partial derivatives. For a function u(x,t), the heat equation u_t = κu_xx describes diffusion, while the wave equation u_tt = c²u_xx describes propagation. The order is the highest derivative appearing, and the equation is linear when the unknown function and its derivatives occur only to the first power and are not multiplied together. Otherwise it is nonlinear. Coefficients may be constant or variable, and the domain may be a line, region, surface, or higher-dimensional space.¹

Unlike an ordinary differential equation, which differentiates with respect to one independent variable, a partial differential equation can encode spatial variation alongside temporal evolution. A complete problem normally includes initial conditions, boundary conditions, or both; these determine whether a solution exists, is unique, and depends continuously on the data—the three requirements often called well-posedness.³

Second-order equations are commonly classified as elliptic, parabolic, or hyperbolic according to the quadratic form made from their highest-order terms. The Laplace equation is elliptic and models equilibrium fields; the heat equation is parabolic and represents irreversible smoothing; the wave equation is hyperbolic and permits finite-speed propagation. This classification predicts characteristic behavior and guides the choice of boundary or initial data.²

Solutions may be obtained by separation of variables, Fourier analysis, transform methods, characteristic curves, Green functions, or variational formulations. In difficult problems, classical derivatives may not exist everywhere, so analysts use a weak solution, which satisfies an integral identity against suitable test functions. Weak formulations are central to modern existence theory and to computational methods such as the finite element method.³ Exact formulas are often less important than proving regularity, uniqueness, stability, or long-time behavior.

Partial differential equations are the main language for continuum models in science and engineering. They govern groundwater flow, elasticity, quantum mechanics, atmospheric dynamics, electromagnetic fields, and biological transport. The Navier–Stokes equations describe viscous fluid motion, but fundamental questions about smoothness and existence in three dimensions remain unresolved and form one of the Clay Mathematics Institute’s Millennium Prize Problems.⁴

Because realistic domains and coefficients rarely yield closed-form solutions, numerical analysis replaces the continuous problem with a discrete one. Finite differences approximate derivatives on a grid; finite elements approximate the solution in a space of basis functions; and finite volumes enforce conservation over small control regions. A useful numerical method must balance consistency, stability, convergence, computational cost, and preservation of physical properties such as positivity or mass conservation.

Partial differential equation theory extends well beyond the familiar three prototype equations. Degenerate equations can lose their usual classification at selected points, while free-boundary problems require the unknown solution and the boundary of its domain to be determined together. Singular solutions may develop shocks, interfaces, or concentrated sources even when the initial data are smooth. In nonlinear equations, weak solutions can exist without being unique unless an additional admissibility or entropy condition selects the physically relevant one.³

The subject also connects deeply with geometry and probability. The heat equation is related to Brownian motion and the smoothing of geometric shapes, while elliptic equations describe curvature and harmonic functions on manifolds. Special-function theory supplies many classical solutions through Bessel, Legendre, and related functions, whose properties are catalogued in the NIST Digital Library of Mathematical Functions.⁵ These links make the field both a modeling discipline and a source of general principles about propagation, regularity, and scale.

Boundary condition

A prescribed value, derivative, or relation imposed on the boundary of the domain.

Characteristic

A curve or surface along which information or singularities may propagate, especially in hyperbolic equations.

Elliptic equation

A class generally associated with equilibrium, spatial smoothing, and boundary-value problems.

Weak solution

A function satisfying a differential equation in an integral sense when classical derivatives may not exist.

Well-posed problem

A problem with existence, uniqueness, and continuous dependence of the solution on the data.