A differential equation relates an unknown function to its own derivatives — describing not the function's value directly, but the rule governing how it changes. Since virtually every physical law describes a rate of change (velocity, growth, decay, cooling, orbital motion), differential equations are the primary mathematical language connecting calculus to the real, dynamic world.
Most equations you learn in school ask you to find a number: solve for x. A differential equation asks something more ambitious: find an entire function, given a rule about how fast that function changes. For example: "a population grows at a rate proportional to its current size" is a differential equation — it doesn't tell you the population directly, but tells you the rule its growth must obey.
Solving a differential equation means finding the actual function that obeys the given rate rule — not just a number, but an entire curve. Some differential equations have neat, explicit formula solutions. Many others, especially those describing complex real-world systems, cannot be solved by any formula at all, and must instead be approximated using numerical computer methods.
An ordinary differential equation (ODE) involves derivatives with respect to a single variable (usually time). A partial differential equation (PDE) involves derivatives with respect to multiple variables simultaneously (like both space and time) — describing phenomena that vary across both dimensions, such as heat spreading through a metal rod or a wave rippling across a pond.
The order is the highest derivative appearing in the equation. dy/dx = ky (exponential growth/decay) is first-order. Newton's Second Law, F=m(d²x/dt²), is second-order, since it involves acceleration (the second derivative of position).
The simplest solvable class is separable equations, where the equation can be algebraically rearranged so all the y-terms are on one side and all the x-terms on the other, then both sides integrated separately. For dy/dx = ky: separating gives dy/y = k dx, integrating gives ln|y| = kx + C, and solving gives the exponential solution y = Ae^(kx) — the formula underlying both population growth and radioactive decay.
A differential equation alone typically has infinitely many possible solutions (differing by a constant, like the "A" above). An initial condition — specifying the function's exact value at one particular point — pins down the unique specific solution matching your actual real-world starting scenario. Together, the equation plus the initial condition form an initial value problem.
A differential equation is linear if the unknown function and its derivatives appear only to the first power, with no products of the function with itself or its derivatives. Linear equations are generally far more tractable — entire systematic theories exist for solving them exactly. Nonlinear equations, which govern many of the most interesting real-world phenomena (turbulence, weather, population interactions), are typically far harder, and many have no known closed-form solution at all.
Before attempting to solve a differential equation, mathematicians ask a more basic question: does a solution even exist, and if so, is it unique? The Picard–Lindelöf theorem (also called the Cauchy-Lipschitz theorem) guarantees that, under a technical smoothness condition (Lipschitz continuity) on the right-hand side of the equation, an initial value problem has exactly one solution, at least locally near the starting point. This theoretical guarantee is essential — without it, numerical methods might be computing an approximation to a solution that doesn't actually exist, or might be capturing only one of several equally valid solutions.
Most differential equations arising from real-world applications cannot be solved by any explicit formula. Instead, numerical methods approximate the solution step by step. The simplest, Euler's method (1768), takes small steps forward using the equation's instantaneous rate of change at each point. More sophisticated methods (like the widely used fourth-order Runge-Kutta method) achieve dramatically better accuracy for the same computational effort, and modern scientific computing relies heavily on such numerical integration techniques for everything from weather forecasting to spacecraft trajectory planning.
Some nonlinear differential equation systems exhibit chaotic behaviour — extreme sensitivity where minuscule differences in starting conditions lead to wildly diverging outcomes over time, even though the underlying equations themselves are entirely deterministic. Edward Lorenz's 1963 study of a simplified atmospheric convection model (leading to the popularised "butterfly effect") is the canonical example, and chaos theory has since become essential to understanding weather prediction limits, fluid turbulence, and numerous other complex dynamical systems.
The Navier-Stokes equations — a system of nonlinear partial differential equations describing fluid flow — are used constantly and successfully throughout engineering and physics, yet a fundamental mathematical question about them remains unresolved: does a smooth, well-behaved solution always exist for all time, given smooth initial conditions in three dimensions? This is one of the seven Clay Millennium Prize Problems (see the Hilbert's 23 Problems entry for context on such problem lists), carrying a $1 million reward, and remains completely open as of 2026 — a striking illustration of how a set of equations can be used with enormous practical success in engineering while a basic theoretical question about their mathematical behaviour remains unanswered.
Maxwell's equations (electromagnetism), the Schrödinger equation (quantum mechanics), and Einstein's field equations (general relativity) are all differential equations — each, in its own domain, serving as the fundamental dynamical law from which essentially all further physical prediction is derived.